Towards large-scale FAME-based bacterial species identification using machine learning techniques.

PubMed

Slabbinck, Bram; De Baets, Bernard; Dawyndt, Peter; De Vos, Paul

2009-05-01

In the last decade, bacterial taxonomy witnessed a huge expansion. The swift pace of bacterial species (re-)definitions has a serious impact on the accuracy and completeness of first-line identification methods. Consequently, back-end identification libraries need to be synchronized with the List of Prokaryotic names with Standing in Nomenclature. In this study, we focus on bacterial fatty acid methyl ester (FAME) profiling as a broadly used first-line identification method. From the BAME@LMG database, we have selected FAME profiles of individual strains belonging to the genera Bacillus, Paenibacillus and Pseudomonas. Only those profiles resulting from standard growth conditions have been retained. The corresponding data set covers 74, 44 and 95 validly published bacterial species, respectively, represented by 961, 378 and 1673 standard FAME profiles. Through the application of machine learning techniques in a supervised strategy, different computational models have been built for genus and species identification. Three techniques have been considered: artificial neural networks, random forests and support vector machines. Nearly perfect identification has been achieved at genus level. Notwithstanding the known limited discriminative power of FAME analysis for species identification, the computational models have resulted in good species identification results for the three genera. For Bacillus, Paenibacillus and Pseudomonas, random forests have resulted in sensitivity values, respectively, 0.847, 0.901 and 0.708. The random forests models outperform those of the other machine learning techniques. Moreover, our machine learning approach also outperformed the Sherlock MIS (MIDI Inc., Newark, DE, USA). These results show that machine learning proves very useful for FAME-based bacterial species identification. Besides good bacterial identification at species level, speed and ease of taxonomic synchronization are major advantages of this computational species identification strategy.

Clustering Single-Cell Expression Data Using Random Forest Graphs.

PubMed

Pouyan, Maziyar Baran; Nourani, Mehrdad

2017-07-01

Complex tissues such as brain and bone marrow are made up of multiple cell types. As the study of biological tissue structure progresses, the role of cell-type-specific research becomes increasingly important. Novel sequencing technology such as single-cell cytometry provides researchers access to valuable biological data. Applying machine-learning techniques to these high-throughput datasets provides deep insights into the cellular landscape of the tissue where those cells are a part of. In this paper, we propose the use of random-forest-based single-cell profiling, a new machine-learning-based technique, to profile different cell types of intricate tissues using single-cell cytometry data. Our technique utilizes random forests to capture cell marker dependences and model the cellular populations using the cell network concept. This cellular network helps us discover what cell types are in the tissue. Our experimental results on public-domain datasets indicate promising performance and accuracy of our technique in extracting cell populations of complex tissues.

Classification of large-sized hyperspectral imagery using fast machine learning algorithms

NASA Astrophysics Data System (ADS)

Xia, Junshi; Yokoya, Naoto; Iwasaki, Akira

2017-07-01

We present a framework of fast machine learning algorithms in the context of large-sized hyperspectral images classification from the theoretical to a practical viewpoint. In particular, we assess the performance of random forest (RF), rotation forest (RoF), and extreme learning machine (ELM) and the ensembles of RF and ELM. These classifiers are applied to two large-sized hyperspectral images and compared to the support vector machines. To give the quantitative analysis, we pay attention to comparing these methods when working with high input dimensions and a limited/sufficient training set. Moreover, other important issues such as the computational cost and robustness against the noise are also discussed.

Taxi-Out Time Prediction for Departures at Charlotte Airport Using Machine Learning Techniques

NASA Technical Reports Server (NTRS)

Lee, Hanbong; Malik, Waqar; Jung, Yoon C.

2016-01-01

Predicting the taxi-out times of departures accurately is important for improving airport efficiency and takeoff time predictability. In this paper, we attempt to apply machine learning techniques to actual traffic data at Charlotte Douglas International Airport for taxi-out time prediction. To find the key factors affecting aircraft taxi times, surface surveillance data is first analyzed. From this data analysis, several variables, including terminal concourse, spot, runway, departure fix and weight class, are selected for taxi time prediction. Then, various machine learning methods such as linear regression, support vector machines, k-nearest neighbors, random forest, and neural networks model are applied to actual flight data. Different traffic flow and weather conditions at Charlotte airport are also taken into account for more accurate prediction. The taxi-out time prediction results show that linear regression and random forest techniques can provide the most accurate prediction in terms of root-mean-square errors. We also discuss the operational complexity and uncertainties that make it difficult to predict the taxi times accurately.

Computer-Aided Diagnosis for Breast Ultrasound Using Computerized BI-RADS Features and Machine Learning Methods.

PubMed

Shan, Juan; Alam, S Kaisar; Garra, Brian; Zhang, Yingtao; Ahmed, Tahira

2016-04-01

This work identifies effective computable features from the Breast Imaging Reporting and Data System (BI-RADS), to develop a computer-aided diagnosis (CAD) system for breast ultrasound. Computerized features corresponding to ultrasound BI-RADs categories were designed and tested using a database of 283 pathology-proven benign and malignant lesions. Features were selected based on classification performance using a "bottom-up" approach for different machine learning methods, including decision tree, artificial neural network, random forest and support vector machine. Using 10-fold cross-validation on the database of 283 cases, the highest area under the receiver operating characteristic (ROC) curve (AUC) was 0.84 from a support vector machine with 77.7% overall accuracy; the highest overall accuracy, 78.5%, was from a random forest with the AUC 0.83. Lesion margin and orientation were optimum features common to all of the different machine learning methods. These features can be used in CAD systems to help distinguish benign from worrisome lesions. Copyright © 2016 World Federation for Ultrasound in Medicine & Biology. All rights reserved.

Assessing and comparison of different machine learning methods in parent-offspring trios for genotype imputation.

PubMed

Mikhchi, Abbas; Honarvar, Mahmood; Kashan, Nasser Emam Jomeh; Aminafshar, Mehdi

2016-06-21

Genotype imputation is an important tool for prediction of unknown genotypes for both unrelated individuals and parent-offspring trios. Several imputation methods are available and can either employ universal machine learning methods, or deploy algorithms dedicated to infer missing genotypes. In this research the performance of eight machine learning methods: Support Vector Machine, K-Nearest Neighbors, Extreme Learning Machine, Radial Basis Function, Random Forest, AdaBoost, LogitBoost, and TotalBoost compared in terms of the imputation accuracy, computation time and the factors affecting imputation accuracy. The methods employed using real and simulated datasets to impute the un-typed SNPs in parent-offspring trios. The tested methods show that imputation of parent-offspring trios can be accurate. The Random Forest and Support Vector Machine were more accurate than the other machine learning methods. The TotalBoost performed slightly worse than the other methods.The running times were different between methods. The ELM was always most fast algorithm. In case of increasing the sample size, the RBF requires long imputation time.The tested methods in this research can be an alternative for imputation of un-typed SNPs in low missing rate of data. However, it is recommended that other machine learning methods to be used for imputation. Copyright © 2016 Elsevier Ltd. All rights reserved.

Accurate Diabetes Risk Stratification Using Machine Learning: Role of Missing Value and Outliers.

PubMed

Maniruzzaman, Md; Rahman, Md Jahanur; Al-MehediHasan, Md; Suri, Harman S; Abedin, Md Menhazul; El-Baz, Ayman; Suri, Jasjit S

2018-04-10

Diabetes mellitus is a group of metabolic diseases in which blood sugar levels are too high. About 8.8% of the world was diabetic in 2017. It is projected that this will reach nearly 10% by 2045. The major challenge is that when machine learning-based classifiers are applied to such data sets for risk stratification, leads to lower performance. Thus, our objective is to develop an optimized and robust machine learning (ML) system under the assumption that missing values or outliers if replaced by a median configuration will yield higher risk stratification accuracy. This ML-based risk stratification is designed, optimized and evaluated, where: (i) the features are extracted and optimized from the six feature selection techniques (random forest, logistic regression, mutual information, principal component analysis, analysis of variance, and Fisher discriminant ratio) and combined with ten different types of classifiers (linear discriminant analysis, quadratic discriminant analysis, naïve Bayes, Gaussian process classification, support vector machine, artificial neural network, Adaboost, logistic regression, decision tree, and random forest) under the hypothesis that both missing values and outliers when replaced by computed medians will improve the risk stratification accuracy. Pima Indian diabetic dataset (768 patients: 268 diabetic and 500 controls) was used. Our results demonstrate that on replacing the missing values and outliers by group median and median values, respectively and further using the combination of random forest feature selection and random forest classification technique yields an accuracy, sensitivity, specificity, positive predictive value, negative predictive value and area under the curve as: 92.26%, 95.96%, 79.72%, 91.14%, 91.20%, and 0.93, respectively. This is an improvement of 10% over previously developed techniques published in literature. The system was validated for its stability and reliability. RF-based model showed the best performance when outliers are replaced by median values.

Mortality risk prediction in burn injury: Comparison of logistic regression with machine learning approaches.

PubMed

Stylianou, Neophytos; Akbarov, Artur; Kontopantelis, Evangelos; Buchan, Iain; Dunn, Ken W

2015-08-01

Predicting mortality from burn injury has traditionally employed logistic regression models. Alternative machine learning methods have been introduced in some areas of clinical prediction as the necessary software and computational facilities have become accessible. Here we compare logistic regression and machine learning predictions of mortality from burn. An established logistic mortality model was compared to machine learning methods (artificial neural network, support vector machine, random forests and naïve Bayes) using a population-based (England & Wales) case-cohort registry. Predictive evaluation used: area under the receiver operating characteristic curve; sensitivity; specificity; positive predictive value and Youden's index. All methods had comparable discriminatory abilities, similar sensitivities, specificities and positive predictive values. Although some machine learning methods performed marginally better than logistic regression the differences were seldom statistically significant and clinically insubstantial. Random forests were marginally better for high positive predictive value and reasonable sensitivity. Neural networks yielded slightly better prediction overall. Logistic regression gives an optimal mix of performance and interpretability. The established logistic regression model of burn mortality performs well against more complex alternatives. Clinical prediction with a small set of strong, stable, independent predictors is unlikely to gain much from machine learning outside specialist research contexts. Copyright © 2015 Elsevier Ltd and ISBI. All rights reserved.

MLACP: machine-learning-based prediction of anticancer peptides

PubMed Central

Manavalan, Balachandran; Basith, Shaherin; Shin, Tae Hwan; Choi, Sun; Kim, Myeong Ok; Lee, Gwang

2017-01-01

Cancer is the second leading cause of death globally, and use of therapeutic peptides to target and kill cancer cells has received considerable attention in recent years. Identification of anticancer peptides (ACPs) through wet-lab experimentation is expensive and often time consuming; therefore, development of an efficient computational method is essential to identify potential ACP candidates prior to in vitro experimentation. In this study, we developed support vector machine- and random forest-based machine-learning methods for the prediction of ACPs using the features calculated from the amino acid sequence, including amino acid composition, dipeptide composition, atomic composition, and physicochemical properties. We trained our methods using the Tyagi-B dataset and determined the machine parameters by 10-fold cross-validation. Furthermore, we evaluated the performance of our methods on two benchmarking datasets, with our results showing that the random forest-based method outperformed the existing methods with an average accuracy and Matthews correlation coefficient value of 88.7% and 0.78, respectively. To assist the scientific community, we also developed a publicly accessible web server at www.thegleelab.org/MLACP.html. PMID:29100375

Machine learning models in breast cancer survival prediction.

PubMed

Montazeri, Mitra; Montazeri, Mohadeseh; Montazeri, Mahdieh; Beigzadeh, Amin

2016-01-01

Breast cancer is one of the most common cancers with a high mortality rate among women. With the early diagnosis of breast cancer survival will increase from 56% to more than 86%. Therefore, an accurate and reliable system is necessary for the early diagnosis of this cancer. The proposed model is the combination of rules and different machine learning techniques. Machine learning models can help physicians to reduce the number of false decisions. They try to exploit patterns and relationships among a large number of cases and predict the outcome of a disease using historical cases stored in datasets. The objective of this study is to propose a rule-based classification method with machine learning techniques for the prediction of different types of Breast cancer survival. We use a dataset with eight attributes that include the records of 900 patients in which 876 patients (97.3%) and 24 (2.7%) patients were females and males respectively. Naive Bayes (NB), Trees Random Forest (TRF), 1-Nearest Neighbor (1NN), AdaBoost (AD), Support Vector Machine (SVM), RBF Network (RBFN), and Multilayer Perceptron (MLP) machine learning techniques with 10-cross fold technique were used with the proposed model for the prediction of breast cancer survival. The performance of machine learning techniques were evaluated with accuracy, precision, sensitivity, specificity, and area under ROC curve. Out of 900 patients, 803 patients and 97 patients were alive and dead, respectively. In this study, Trees Random Forest (TRF) technique showed better results in comparison to other techniques (NB, 1NN, AD, SVM and RBFN, MLP). The accuracy, sensitivity and the area under ROC curve of TRF are 96%, 96%, 93%, respectively. However, 1NN machine learning technique provided poor performance (accuracy 91%, sensitivity 91% and area under ROC curve 78%). This study demonstrates that Trees Random Forest model (TRF) which is a rule-based classification model was the best model with the highest level of accuracy. Therefore, this model is recommended as a useful tool for breast cancer survival prediction as well as medical decision making.

A Machine Learning and Cross-Validation Approach for the Discrimination of Vegetation Physiognomic Types Using Satellite Based Multispectral and Multitemporal Data.

PubMed

Sharma, Ram C; Hara, Keitarou; Hirayama, Hidetake

2017-01-01

This paper presents the performance and evaluation of a number of machine learning classifiers for the discrimination between the vegetation physiognomic classes using the satellite based time-series of the surface reflectance data. Discrimination of six vegetation physiognomic classes, Evergreen Coniferous Forest, Evergreen Broadleaf Forest, Deciduous Coniferous Forest, Deciduous Broadleaf Forest, Shrubs, and Herbs, was dealt with in the research. Rich-feature data were prepared from time-series of the satellite data for the discrimination and cross-validation of the vegetation physiognomic types using machine learning approach. A set of machine learning experiments comprised of a number of supervised classifiers with different model parameters was conducted to assess how the discrimination of vegetation physiognomic classes varies with classifiers, input features, and ground truth data size. The performance of each experiment was evaluated by using the 10-fold cross-validation method. Experiment using the Random Forests classifier provided highest overall accuracy (0.81) and kappa coefficient (0.78). However, accuracy metrics did not vary much with experiments. Accuracy metrics were found to be very sensitive to input features and size of ground truth data. The results obtained in the research are expected to be useful for improving the vegetation physiognomic mapping in Japan.

Machine learning to predict the occurrence of bisphosphonate-related osteonecrosis of the jaw associated with dental extraction: A preliminary report.

PubMed

Kim, Dong Wook; Kim, Hwiyoung; Nam, Woong; Kim, Hyung Jun; Cha, In-Ho

2018-04-23

The aim of this study was to build and validate five types of machine learning models that can predict the occurrence of BRONJ associated with dental extraction in patients taking bisphosphonates for the management of osteoporosis. A retrospective review of the medical records was conducted to obtain cases and controls for the study. Total 125 patients consisting of 41 cases and 84 controls were selected for the study. Five machine learning prediction algorithms including multivariable logistic regression model, decision tree, support vector machine, artificial neural network, and random forest were implemented. The outputs of these models were compared with each other and also with conventional methods, such as serum CTX level. Area under the receiver operating characteristic (ROC) curve (AUC) was used to compare the results. The performance of machine learning models was significantly superior to conventional statistical methods and single predictors. The random forest model yielded the best performance (AUC = 0.973), followed by artificial neural network (AUC = 0.915), support vector machine (AUC = 0.882), logistic regression (AUC = 0.844), decision tree (AUC = 0.821), drug holiday alone (AUC = 0.810), and CTX level alone (AUC = 0.630). Machine learning methods showed superior performance in predicting BRONJ associated with dental extraction compared to conventional statistical methods using drug holiday and serum CTX level. Machine learning can thus be applied in a wide range of clinical studies. Copyright © 2017. Published by Elsevier Inc.

The influence of negative training set size on machine learning-based virtual screening.

PubMed

Kurczab, Rafał; Smusz, Sabina; Bojarski, Andrzej J

2014-01-01

The paper presents a thorough analysis of the influence of the number of negative training examples on the performance of machine learning methods. The impact of this rather neglected aspect of machine learning methods application was examined for sets containing a fixed number of positive and a varying number of negative examples randomly selected from the ZINC database. An increase in the ratio of positive to negative training instances was found to greatly influence most of the investigated evaluating parameters of ML methods in simulated virtual screening experiments. In a majority of cases, substantial increases in precision and MCC were observed in conjunction with some decreases in hit recall. The analysis of dynamics of those variations let us recommend an optimal composition of training data. The study was performed on several protein targets, 5 machine learning algorithms (SMO, Naïve Bayes, Ibk, J48 and Random Forest) and 2 types of molecular fingerprints (MACCS and CDK FP). The most effective classification was provided by the combination of CDK FP with SMO or Random Forest algorithms. The Naïve Bayes models appeared to be hardly sensitive to changes in the number of negative instances in the training set. In conclusion, the ratio of positive to negative training instances should be taken into account during the preparation of machine learning experiments, as it might significantly influence the performance of particular classifier. What is more, the optimization of negative training set size can be applied as a boosting-like approach in machine learning-based virtual screening.

The influence of negative training set size on machine learning-based virtual screening

PubMed Central

2014-01-01

Background The paper presents a thorough analysis of the influence of the number of negative training examples on the performance of machine learning methods. Results The impact of this rather neglected aspect of machine learning methods application was examined for sets containing a fixed number of positive and a varying number of negative examples randomly selected from the ZINC database. An increase in the ratio of positive to negative training instances was found to greatly influence most of the investigated evaluating parameters of ML methods in simulated virtual screening experiments. In a majority of cases, substantial increases in precision and MCC were observed in conjunction with some decreases in hit recall. The analysis of dynamics of those variations let us recommend an optimal composition of training data. The study was performed on several protein targets, 5 machine learning algorithms (SMO, Naïve Bayes, Ibk, J48 and Random Forest) and 2 types of molecular fingerprints (MACCS and CDK FP). The most effective classification was provided by the combination of CDK FP with SMO or Random Forest algorithms. The Naïve Bayes models appeared to be hardly sensitive to changes in the number of negative instances in the training set. Conclusions In conclusion, the ratio of positive to negative training instances should be taken into account during the preparation of machine learning experiments, as it might significantly influence the performance of particular classifier. What is more, the optimization of negative training set size can be applied as a boosting-like approach in machine learning-based virtual screening. PMID:24976867

Comparison of Random Forest and Parametric Imputation Models for Imputing Missing Data Using MICE: A CALIBER Study

PubMed Central

Shah, Anoop D.; Bartlett, Jonathan W.; Carpenter, James; Nicholas, Owen; Hemingway, Harry

2014-01-01

Multivariate imputation by chained equations (MICE) is commonly used for imputing missing data in epidemiologic research. The “true” imputation model may contain nonlinearities which are not included in default imputation models. Random forest imputation is a machine learning technique which can accommodate nonlinearities and interactions and does not require a particular regression model to be specified. We compared parametric MICE with a random forest-based MICE algorithm in 2 simulation studies. The first study used 1,000 random samples of 2,000 persons drawn from the 10,128 stable angina patients in the CALIBER database (Cardiovascular Disease Research using Linked Bespoke Studies and Electronic Records; 2001–2010) with complete data on all covariates. Variables were artificially made “missing at random,” and the bias and efficiency of parameter estimates obtained using different imputation methods were compared. Both MICE methods produced unbiased estimates of (log) hazard ratios, but random forest was more efficient and produced narrower confidence intervals. The second study used simulated data in which the partially observed variable depended on the fully observed variables in a nonlinear way. Parameter estimates were less biased using random forest MICE, and confidence interval coverage was better. This suggests that random forest imputation may be useful for imputing complex epidemiologic data sets in which some patients have missing data. PMID:24589914

Comparison of random forest and parametric imputation models for imputing missing data using MICE: a CALIBER study.

PubMed

Shah, Anoop D; Bartlett, Jonathan W; Carpenter, James; Nicholas, Owen; Hemingway, Harry

2014-03-15

Multivariate imputation by chained equations (MICE) is commonly used for imputing missing data in epidemiologic research. The "true" imputation model may contain nonlinearities which are not included in default imputation models. Random forest imputation is a machine learning technique which can accommodate nonlinearities and interactions and does not require a particular regression model to be specified. We compared parametric MICE with a random forest-based MICE algorithm in 2 simulation studies. The first study used 1,000 random samples of 2,000 persons drawn from the 10,128 stable angina patients in the CALIBER database (Cardiovascular Disease Research using Linked Bespoke Studies and Electronic Records; 2001-2010) with complete data on all covariates. Variables were artificially made "missing at random," and the bias and efficiency of parameter estimates obtained using different imputation methods were compared. Both MICE methods produced unbiased estimates of (log) hazard ratios, but random forest was more efficient and produced narrower confidence intervals. The second study used simulated data in which the partially observed variable depended on the fully observed variables in a nonlinear way. Parameter estimates were less biased using random forest MICE, and confidence interval coverage was better. This suggests that random forest imputation may be useful for imputing complex epidemiologic data sets in which some patients have missing data.

Studies of the DIII-D disruption database using Machine Learning algorithms

NASA Astrophysics Data System (ADS)

Rea, Cristina; Granetz, Robert; Meneghini, Orso

2017-10-01

A Random Forests Machine Learning algorithm, trained on a large database of both disruptive and non-disruptive DIII-D discharges, predicts disruptive behavior in DIII-D with about 90% of accuracy. Several algorithms have been tested and Random Forests was found superior in performances for this particular task. Over 40 plasma parameters are included in the database, with data for each of the parameters taken from 500k time slices. We focused on a subset of non-dimensional plasma parameters, deemed to be good predictors based on physics considerations. Both binary (disruptive/non-disruptive) and multi-label (label based on the elapsed time before disruption) classification problems are investigated. The Random Forests algorithm provides insight on the available dataset by ranking the relative importance of the input features. It is found that q95 and Greenwald density fraction (n/nG) are the most relevant parameters for discriminating between DIII-D disruptive and non-disruptive discharges. A comparison with the Gradient Boosted Trees algorithm is shown and the first results coming from the application of regression algorithms are presented. Work supported by the US Department of Energy under DE-FC02-04ER54698, DE-SC0014264 and DE-FG02-95ER54309.

Applying a weighted random forests method to extract karst sinkholes from LiDAR data

NASA Astrophysics Data System (ADS)

Zhu, Junfeng; Pierskalla, William P.

2016-02-01

Detailed mapping of sinkholes provides critical information for mitigating sinkhole hazards and understanding groundwater and surface water interactions in karst terrains. LiDAR (Light Detection and Ranging) measures the earth's surface in high-resolution and high-density and has shown great potentials to drastically improve locating and delineating sinkholes. However, processing LiDAR data to extract sinkholes requires separating sinkholes from other depressions, which can be laborious because of the sheer number of the depressions commonly generated from LiDAR data. In this study, we applied the random forests, a machine learning method, to automatically separate sinkholes from other depressions in a karst region in central Kentucky. The sinkhole-extraction random forest was grown on a training dataset built from an area where LiDAR-derived depressions were manually classified through a visual inspection and field verification process. Based on the geometry of depressions, as well as natural and human factors related to sinkholes, 11 parameters were selected as predictive variables to form the dataset. Because the training dataset was imbalanced with the majority of depressions being non-sinkholes, a weighted random forests method was used to improve the accuracy of predicting sinkholes. The weighted random forest achieved an average accuracy of 89.95% for the training dataset, demonstrating that the random forest can be an effective sinkhole classifier. Testing of the random forest in another area, however, resulted in moderate success with an average accuracy rate of 73.96%. This study suggests that an automatic sinkhole extraction procedure like the random forest classifier can significantly reduce time and labor costs and makes its more tractable to map sinkholes using LiDAR data for large areas. However, the random forests method cannot totally replace manual procedures, such as visual inspection and field verification.

Development of machine learning models for diagnosis of glaucoma.

PubMed

Kim, Seong Jae; Cho, Kyong Jin; Oh, Sejong

2017-01-01

The study aimed to develop machine learning models that have strong prediction power and interpretability for diagnosis of glaucoma based on retinal nerve fiber layer (RNFL) thickness and visual field (VF). We collected various candidate features from the examination of retinal nerve fiber layer (RNFL) thickness and visual field (VF). We also developed synthesized features from original features. We then selected the best features proper for classification (diagnosis) through feature evaluation. We used 100 cases of data as a test dataset and 399 cases of data as a training and validation dataset. To develop the glaucoma prediction model, we considered four machine learning algorithms: C5.0, random forest (RF), support vector machine (SVM), and k-nearest neighbor (KNN). We repeatedly composed a learning model using the training dataset and evaluated it by using the validation dataset. Finally, we got the best learning model that produces the highest validation accuracy. We analyzed quality of the models using several measures. The random forest model shows best performance and C5.0, SVM, and KNN models show similar accuracy. In the random forest model, the classification accuracy is 0.98, sensitivity is 0.983, specificity is 0.975, and AUC is 0.979. The developed prediction models show high accuracy, sensitivity, specificity, and AUC in classifying among glaucoma and healthy eyes. It will be used for predicting glaucoma against unknown examination records. Clinicians may reference the prediction results and be able to make better decisions. We may combine multiple learning models to increase prediction accuracy. The C5.0 model includes decision rules for prediction. It can be used to explain the reasons for specific predictions.

Prediction of Return-to-original-work after an Industrial Accident Using Machine Learning and Comparison of Techniques

PubMed Central

2018-01-01

Background Many studies have tried to develop predictors for return-to-work (RTW). However, since complex factors have been demonstrated to predict RTW, it is difficult to use them practically. This study investigated whether factors used in previous studies could predict whether an individual had returned to his/her original work by four years after termination of the worker's recovery period. Methods An initial logistic regression analysis of 1,567 participants of the fourth Panel Study of Worker's Compensation Insurance yielded odds ratios. The participants were divided into two subsets, a training dataset and a test dataset. Using the training dataset, logistic regression, decision tree, random forest, and support vector machine models were established, and important variables of each model were identified. The predictive abilities of the different models were compared. Results The analysis showed that only earned income and company-related factors significantly affected return-to-original-work (RTOW). The random forest model showed the best accuracy among the tested machine learning models; however, the difference was not prominent. Conclusion It is possible to predict a worker's probability of RTOW using machine learning techniques with moderate accuracy. PMID:29736160

Identifying Active Travel Behaviors in Challenging Environments Using GPS, Accelerometers, and Machine Learning Algorithms.

PubMed

Ellis, Katherine; Godbole, Suneeta; Marshall, Simon; Lanckriet, Gert; Staudenmayer, John; Kerr, Jacqueline

2014-01-01

Active travel is an important area in physical activity research, but objective measurement of active travel is still difficult. Automated methods to measure travel behaviors will improve research in this area. In this paper, we present a supervised machine learning method for transportation mode prediction from global positioning system (GPS) and accelerometer data. We collected a dataset of about 150 h of GPS and accelerometer data from two research assistants following a protocol of prescribed trips consisting of five activities: bicycling, riding in a vehicle, walking, sitting, and standing. We extracted 49 features from 1-min windows of this data. We compared the performance of several machine learning algorithms and chose a random forest algorithm to classify the transportation mode. We used a moving average output filter to smooth the output predictions over time. The random forest algorithm achieved 89.8% cross-validated accuracy on this dataset. Adding the moving average filter to smooth output predictions increased the cross-validated accuracy to 91.9%. Machine learning methods are a viable approach for automating measurement of active travel, particularly for measuring travel activities that traditional accelerometer data processing methods misclassify, such as bicycling and vehicle travel.

A machine learning system to improve heart failure patient assistance.

PubMed

Guidi, Gabriele; Pettenati, Maria Chiara; Melillo, Paolo; Iadanza, Ernesto

2014-11-01

In this paper, we present a clinical decision support system (CDSS) for the analysis of heart failure (HF) patients, providing various outputs such as an HF severity evaluation, HF-type prediction, as well as a management interface that compares the different patients' follow-ups. The whole system is composed of a part of intelligent core and of an HF special-purpose management tool also providing the function to act as interface for the artificial intelligence training and use. To implement the smart intelligent functions, we adopted a machine learning approach. In this paper, we compare the performance of a neural network (NN), a support vector machine, a system with fuzzy rules genetically produced, and a classification and regression tree and its direct evolution, which is the random forest, in analyzing our database. Best performances in both HF severity evaluation and HF-type prediction functions are obtained by using the random forest algorithm. The management tool allows the cardiologist to populate a "supervised database" suitable for machine learning during his or her regular outpatient consultations. The idea comes from the fact that in literature there are a few databases of this type, and they are not scalable to our case.

Prediction of drug synergy in cancer using ensemble-based machine learning techniques

NASA Astrophysics Data System (ADS)

Singh, Harpreet; Rana, Prashant Singh; Singh, Urvinder

2018-04-01

Drug synergy prediction plays a significant role in the medical field for inhibiting specific cancer agents. It can be developed as a pre-processing tool for therapeutic successes. Examination of different drug-drug interaction can be done by drug synergy score. It needs efficient regression-based machine learning approaches to minimize the prediction errors. Numerous machine learning techniques such as neural networks, support vector machines, random forests, LASSO, Elastic Nets, etc., have been used in the past to realize requirement as mentioned above. However, these techniques individually do not provide significant accuracy in drug synergy score. Therefore, the primary objective of this paper is to design a neuro-fuzzy-based ensembling approach. To achieve this, nine well-known machine learning techniques have been implemented by considering the drug synergy data. Based on the accuracy of each model, four techniques with high accuracy are selected to develop ensemble-based machine learning model. These models are Random forest, Fuzzy Rules Using Genetic Cooperative-Competitive Learning method (GFS.GCCL), Adaptive-Network-Based Fuzzy Inference System (ANFIS) and Dynamic Evolving Neural-Fuzzy Inference System method (DENFIS). Ensembling is achieved by evaluating the biased weighted aggregation (i.e. adding more weights to the model with a higher prediction score) of predicted data by selected models. The proposed and existing machine learning techniques have been evaluated on drug synergy score data. The comparative analysis reveals that the proposed method outperforms others in terms of accuracy, root mean square error and coefficient of correlation.

A machine learning method to estimate PM2.5 concentrations across China with remote sensing, meteorological and land use information.

PubMed

Chen, Gongbo; Li, Shanshan; Knibbs, Luke D; Hamm, N A S; Cao, Wei; Li, Tiantian; Guo, Jianping; Ren, Hongyan; Abramson, Michael J; Guo, Yuming

2018-09-15

Machine learning algorithms have very high predictive ability. However, no study has used machine learning to estimate historical concentrations of PM 2.5 (particulate matter with aerodynamic diameter ≤ 2.5 μm) at daily time scale in China at a national level. To estimate daily concentrations of PM 2.5 across China during 2005-2016. Daily ground-level PM 2.5 data were obtained from 1479 stations across China during 2014-2016. Data on aerosol optical depth (AOD), meteorological conditions and other predictors were downloaded. A random forests model (non-parametric machine learning algorithms) and two traditional regression models were developed to estimate ground-level PM 2.5 concentrations. The best-fit model was then utilized to estimate the daily concentrations of PM 2.5 across China with a resolution of 0.1° (≈10 km) during 2005-2016. The daily random forests model showed much higher predictive accuracy than the other two traditional regression models, explaining the majority of spatial variability in daily PM 2.5 [10-fold cross-validation (CV) R 2  = 83%, root mean squared prediction error (RMSE) = 28.1 μg/m 3 ]. At the monthly and annual time-scale, the explained variability of average PM 2.5 increased up to 86% (RMSE = 10.7 μg/m 3 and 6.9 μg/m 3 , respectively). Taking advantage of a novel application of modeling framework and the most recent ground-level PM 2.5 observations, the machine learning method showed higher predictive ability than previous studies. Random forests approach can be used to estimate historical exposure to PM 2.5 in China with high accuracy. Copyright © 2018 Elsevier B.V. All rights reserved.

The use of single-date MODIS imagery for estimating large-scale urban impervious surface fraction with spectral mixture analysis and machine learning techniques

NASA Astrophysics Data System (ADS)

Deng, Chengbin; Wu, Changshan

2013-12-01

Urban impervious surface information is essential for urban and environmental applications at the regional/national scales. As a popular image processing technique, spectral mixture analysis (SMA) has rarely been applied to coarse-resolution imagery due to the difficulty of deriving endmember spectra using traditional endmember selection methods, particularly within heterogeneous urban environments. To address this problem, we derived endmember signatures through a least squares solution (LSS) technique with known abundances of sample pixels, and integrated these endmember signatures into SMA for mapping large-scale impervious surface fraction. In addition, with the same sample set, we carried out objective comparative analyses among SMA (i.e. fully constrained and unconstrained SMA) and machine learning (i.e. Cubist regression tree and Random Forests) techniques. Analysis of results suggests three major conclusions. First, with the extrapolated endmember spectra from stratified random training samples, the SMA approaches performed relatively well, as indicated by small MAE values. Second, Random Forests yields more reliable results than Cubist regression tree, and its accuracy is improved with increased sample sizes. Finally, comparative analyses suggest a tentative guide for selecting an optimal approach for large-scale fractional imperviousness estimation: unconstrained SMA might be a favorable option with a small number of samples, while Random Forests might be preferred if a large number of samples are available.

Machine Learning Predictions of a Multiresolution Climate Model Ensemble

NASA Astrophysics Data System (ADS)

Anderson, Gemma J.; Lucas, Donald D.

2018-05-01

Statistical models of high-resolution climate models are useful for many purposes, including sensitivity and uncertainty analyses, but building them can be computationally prohibitive. We generated a unique multiresolution perturbed parameter ensemble of a global climate model. We use a novel application of a machine learning technique known as random forests to train a statistical model on the ensemble to make high-resolution model predictions of two important quantities: global mean top-of-atmosphere energy flux and precipitation. The random forests leverage cheaper low-resolution simulations, greatly reducing the number of high-resolution simulations required to train the statistical model. We demonstrate that high-resolution predictions of these quantities can be obtained by training on an ensemble that includes only a small number of high-resolution simulations. We also find that global annually averaged precipitation is more sensitive to resolution changes than to any of the model parameters considered.

Comparison of Models for the Prediction of Medical Costs of Spinal Fusion in Taiwan Diagnosis-Related Groups by Machine Learning Algorithms.

PubMed

Kuo, Ching-Yen; Yu, Liang-Chin; Chen, Hou-Chaung; Chan, Chien-Lung

2018-01-01

The aims of this study were to compare the performance of machine learning methods for the prediction of the medical costs associated with spinal fusion in terms of profit or loss in Taiwan Diagnosis-Related Groups (Tw-DRGs) and to apply these methods to explore the important factors associated with the medical costs of spinal fusion. A data set was obtained from a regional hospital in Taoyuan city in Taiwan, which contained data from 2010 to 2013 on patients of Tw-DRG49702 (posterior and other spinal fusion without complications or comorbidities). Naïve-Bayesian, support vector machines, logistic regression, C4.5 decision tree, and random forest methods were employed for prediction using WEKA 3.8.1. Five hundred thirty-two cases were categorized as belonging to the Tw-DRG49702 group. The mean medical cost was US $4,549.7, and the mean age of the patients was 62.4 years. The mean length of stay was 9.3 days. The length of stay was an important variable in terms of determining medical costs for patients undergoing spinal fusion. The random forest method had the best predictive performance in comparison to the other methods, achieving an accuracy of 84.30%, a sensitivity of 71.4%, a specificity of 92.2%, and an AUC of 0.904. Our study demonstrated that the random forest model can be employed to predict the medical costs of Tw-DRG49702, and could inform hospital strategy in terms of increasing the financial management efficiency of this operation.

The Past, Present and Future of the Meteorological Phenomena Identification Near the Ground (mPING) Project

NASA Astrophysics Data System (ADS)

Elmore, K. L.

2016-12-01

The Metorological Phenomemna Identification NeartheGround (mPING) project is an example of a crowd-sourced, citizen science effort to gather data of sufficeint quality and quantity needed by new post processing methods that use machine learning. Transportation and infrastructure are particularly sensitive to precipitation type in winter weather. We extract attributes from operational numerical forecast models and use them in a random forest to generate forecast winter precipitation types. We find that random forests applied to forecast soundings are effective at generating skillful forecasts of surface ptype with consideralbly more skill than the current algorithms, especuially for ice pellets and freezing rain. We also find that three very different forecast models yuield similar overall results, showing that random forests are able to extract essentially equivalent information from different forecast models. We also show that the random forest for each model, and each profile type is unique to the particular forecast model and that the random forests developed using a particular model suffer significant degradation when given attributes derived from a different model. This implies that no single algorithm can perform well across all forecast models. Clearly, random forests extract information unavailable to "physically based" methods because the physical information in the models does not appear as we expect. One intersting result is that results from the classic "warm nose" sounding profile are, by far, the most sensitive to the particular forecast model, but this profile is also the one for which random forests are most skillful. Finally, a method for calibrarting probabilties for each different ptype using multinomial logistic regression is shown.

Statistical-learning strategies generate only modestly performing predictive models for urinary symptoms following external beam radiotherapy of the prostate: A comparison of conventional and machine-learning methods

DOE Office of Scientific and Technical Information (OSTI.GOV)

Yahya, Noorazrul, E-mail: noorazrul.yahya@research.uwa.edu.au; Ebert, Martin A.; Bulsara, Max

Purpose: Given the paucity of available data concerning radiotherapy-induced urinary toxicity, it is important to ensure derivation of the most robust models with superior predictive performance. This work explores multiple statistical-learning strategies for prediction of urinary symptoms following external beam radiotherapy of the prostate. Methods: The performance of logistic regression, elastic-net, support-vector machine, random forest, neural network, and multivariate adaptive regression splines (MARS) to predict urinary symptoms was analyzed using data from 754 participants accrued by TROG03.04-RADAR. Predictive features included dose-surface data, comorbidities, and medication-intake. Four symptoms were analyzed: dysuria, haematuria, incontinence, and frequency, each with three definitions (grade ≥more » 1, grade ≥ 2 and longitudinal) with event rate between 2.3% and 76.1%. Repeated cross-validations producing matched models were implemented. A synthetic minority oversampling technique was utilized in endpoints with rare events. Parameter optimization was performed on the training data. Area under the receiver operating characteristic curve (AUROC) was used to compare performance using sample size to detect differences of ≥0.05 at the 95% confidence level. Results: Logistic regression, elastic-net, random forest, MARS, and support-vector machine were the highest-performing statistical-learning strategies in 3, 3, 3, 2, and 1 endpoints, respectively. Logistic regression, MARS, elastic-net, random forest, neural network, and support-vector machine were the best, or were not significantly worse than the best, in 7, 7, 5, 5, 3, and 1 endpoints. The best-performing statistical model was for dysuria grade ≥ 1 with AUROC ± standard deviation of 0.649 ± 0.074 using MARS. For longitudinal frequency and dysuria grade ≥ 1, all strategies produced AUROC>0.6 while all haematuria endpoints and longitudinal incontinence models produced AUROC<0.6. Conclusions: Logistic regression and MARS were most likely to be the best-performing strategy for the prediction of urinary symptoms with elastic-net and random forest producing competitive results. The predictive power of the models was modest and endpoint-dependent. New features, including spatial dose maps, may be necessary to achieve better models.« less

Identifying Active Travel Behaviors in Challenging Environments Using GPS, Accelerometers, and Machine Learning Algorithms

PubMed Central

Ellis, Katherine; Godbole, Suneeta; Marshall, Simon; Lanckriet, Gert; Staudenmayer, John; Kerr, Jacqueline

2014-01-01

Background: Active travel is an important area in physical activity research, but objective measurement of active travel is still difficult. Automated methods to measure travel behaviors will improve research in this area. In this paper, we present a supervised machine learning method for transportation mode prediction from global positioning system (GPS) and accelerometer data. Methods: We collected a dataset of about 150 h of GPS and accelerometer data from two research assistants following a protocol of prescribed trips consisting of five activities: bicycling, riding in a vehicle, walking, sitting, and standing. We extracted 49 features from 1-min windows of this data. We compared the performance of several machine learning algorithms and chose a random forest algorithm to classify the transportation mode. We used a moving average output filter to smooth the output predictions over time. Results: The random forest algorithm achieved 89.8% cross-validated accuracy on this dataset. Adding the moving average filter to smooth output predictions increased the cross-validated accuracy to 91.9%. Conclusion: Machine learning methods are a viable approach for automating measurement of active travel, particularly for measuring travel activities that traditional accelerometer data processing methods misclassify, such as bicycling and vehicle travel. PMID:24795875

Application of Machine Learning Approaches for Classifying Sitting Posture Based on Force and Acceleration Sensors.

PubMed

Zemp, Roland; Tanadini, Matteo; Plüss, Stefan; Schnüriger, Karin; Singh, Navrag B; Taylor, William R; Lorenzetti, Silvio

2016-01-01

Occupational musculoskeletal disorders, particularly chronic low back pain (LBP), are ubiquitous due to prolonged static sitting or nonergonomic sitting positions. Therefore, the aim of this study was to develop an instrumented chair with force and acceleration sensors to determine the accuracy of automatically identifying the user's sitting position by applying five different machine learning methods (Support Vector Machines, Multinomial Regression, Boosting, Neural Networks, and Random Forest). Forty-one subjects were requested to sit four times in seven different prescribed sitting positions (total 1148 samples). Sixteen force sensor values and the backrest angle were used as the explanatory variables (features) for the classification. The different classification methods were compared by means of a Leave-One-Out cross-validation approach. The best performance was achieved using the Random Forest classification algorithm, producing a mean classification accuracy of 90.9% for subjects with which the algorithm was not familiar. The classification accuracy varied between 81% and 98% for the seven different sitting positions. The present study showed the possibility of accurately classifying different sitting positions by means of the introduced instrumented office chair combined with machine learning analyses. The use of such novel approaches for the accurate assessment of chair usage could offer insights into the relationships between sitting position, sitting behaviour, and the occurrence of musculoskeletal disorders.

CW-SSIM kernel based random forest for image classification

NASA Astrophysics Data System (ADS)

Fan, Guangzhe; Wang, Zhou; Wang, Jiheng

2010-07-01

Complex wavelet structural similarity (CW-SSIM) index has been proposed as a powerful image similarity metric that is robust to translation, scaling and rotation of images, but how to employ it in image classification applications has not been deeply investigated. In this paper, we incorporate CW-SSIM as a kernel function into a random forest learning algorithm. This leads to a novel image classification approach that does not require a feature extraction or dimension reduction stage at the front end. We use hand-written digit recognition as an example to demonstrate our algorithm. We compare the performance of the proposed approach with random forest learning based on other kernels, including the widely adopted Gaussian and the inner product kernels. Empirical evidences show that the proposed method is superior in its classification power. We also compared our proposed approach with the direct random forest method without kernel and the popular kernel-learning method support vector machine. Our test results based on both simulated and realworld data suggest that the proposed approach works superior to traditional methods without the feature selection procedure.

Source localization in an ocean waveguide using supervised machine learning.

PubMed

Niu, Haiqiang; Reeves, Emma; Gerstoft, Peter

2017-09-01

Source localization in ocean acoustics is posed as a machine learning problem in which data-driven methods learn source ranges directly from observed acoustic data. The pressure received by a vertical linear array is preprocessed by constructing a normalized sample covariance matrix and used as the input for three machine learning methods: feed-forward neural networks (FNN), support vector machines (SVM), and random forests (RF). The range estimation problem is solved both as a classification problem and as a regression problem by these three machine learning algorithms. The results of range estimation for the Noise09 experiment are compared for FNN, SVM, RF, and conventional matched-field processing and demonstrate the potential of machine learning for underwater source localization.

Comparison of the Predictive Performance and Interpretability of Random Forest and Linear Models on Benchmark Data Sets.

PubMed

Marchese Robinson, Richard L; Palczewska, Anna; Palczewski, Jan; Kidley, Nathan

2017-08-28

The ability to interpret the predictions made by quantitative structure-activity relationships (QSARs) offers a number of advantages. While QSARs built using nonlinear modeling approaches, such as the popular Random Forest algorithm, might sometimes be more predictive than those built using linear modeling approaches, their predictions have been perceived as difficult to interpret. However, a growing number of approaches have been proposed for interpreting nonlinear QSAR models in general and Random Forest in particular. In the current work, we compare the performance of Random Forest to those of two widely used linear modeling approaches: linear Support Vector Machines (SVMs) (or Support Vector Regression (SVR)) and partial least-squares (PLS). We compare their performance in terms of their predictivity as well as the chemical interpretability of the predictions using novel scoring schemes for assessing heat map images of substructural contributions. We critically assess different approaches for interpreting Random Forest models as well as for obtaining predictions from the forest. We assess the models on a large number of widely employed public-domain benchmark data sets corresponding to regression and binary classification problems of relevance to hit identification and toxicology. We conclude that Random Forest typically yields comparable or possibly better predictive performance than the linear modeling approaches and that its predictions may also be interpreted in a chemically and biologically meaningful way. In contrast to earlier work looking at interpretation of nonlinear QSAR models, we directly compare two methodologically distinct approaches for interpreting Random Forest models. The approaches for interpreting Random Forest assessed in our article were implemented using open-source programs that we have made available to the community. These programs are the rfFC package ( https://r-forge.r-project.org/R/?group_id=1725 ) for the R statistical programming language and the Python program HeatMapWrapper [ https://doi.org/10.5281/zenodo.495163 ] for heat map generation.

Integrating support vector machines and random forests to classify crops in time series of Worldview-2 images

NASA Astrophysics Data System (ADS)

Zafari, A.; Zurita-Milla, R.; Izquierdo-Verdiguier, E.

2017-10-01

Crop maps are essential inputs for the agricultural planning done at various governmental and agribusinesses agencies. Remote sensing offers timely and costs efficient technologies to identify and map crop types over large areas. Among the plethora of classification methods, Support Vector Machine (SVM) and Random Forest (RF) are widely used because of their proven performance. In this work, we study the synergic use of both methods by introducing a random forest kernel (RFK) in an SVM classifier. A time series of multispectral WorldView-2 images acquired over Mali (West Africa) in 2014 was used to develop our case study. Ground truth containing five common crop classes (cotton, maize, millet, peanut, and sorghum) were collected at 45 farms and used to train and test the classifiers. An SVM with the standard Radial Basis Function (RBF) kernel, a RF, and an SVM-RFK were trained and tested over 10 random training and test subsets generated from the ground data. Results show that the newly proposed SVM-RFK classifier can compete with both RF and SVM-RBF. The overall accuracies based on the spectral bands only are of 83, 82 and 83% respectively. Adding vegetation indices to the analysis result in the classification accuracy of 82, 81 and 84% for SVM-RFK, RF, and SVM-RBF respectively. Overall, it can be observed that the newly tested RFK can compete with SVM-RBF and RF classifiers in terms of classification accuracy.

Performance of thigh-mounted triaxial accelerometer algorithms in objective quantification of sedentary behaviour and physical activity in older adults

PubMed Central

Verschueren, Sabine M. P.; Degens, Hans; Morse, Christopher I.; Onambélé, Gladys L.

2017-01-01

Accurate monitoring of sedentary behaviour and physical activity is key to investigate their exact role in healthy ageing. To date, accelerometers using cut-off point models are most preferred for this, however, machine learning seems a highly promising future alternative. Hence, the current study compared between cut-off point and machine learning algorithms, for optimal quantification of sedentary behaviour and physical activity intensities in the elderly. Thus, in a heterogeneous sample of forty participants (aged ≥60 years, 50% female) energy expenditure during laboratory-based activities (ranging from sedentary behaviour through to moderate-to-vigorous physical activity) was estimated by indirect calorimetry, whilst wearing triaxial thigh-mounted accelerometers. Three cut-off point algorithms and a Random Forest machine learning model were developed and cross-validated using the collected data. Detailed analyses were performed to check algorithm robustness, and examine and benchmark both overall and participant-specific balanced accuracies. This revealed that the four models can at least be used to confidently monitor sedentary behaviour and moderate-to-vigorous physical activity. Nevertheless, the machine learning algorithm outperformed the cut-off point models by being robust for all individual’s physiological and non-physiological characteristics and showing more performance of an acceptable level over the whole range of physical activity intensities. Therefore, we propose that Random Forest machine learning may be optimal for objective assessment of sedentary behaviour and physical activity in older adults using thigh-mounted triaxial accelerometry. PMID:29155839

Performance of thigh-mounted triaxial accelerometer algorithms in objective quantification of sedentary behaviour and physical activity in older adults.

PubMed

Wullems, Jorgen A; Verschueren, Sabine M P; Degens, Hans; Morse, Christopher I; Onambélé, Gladys L

2017-01-01

Accurate monitoring of sedentary behaviour and physical activity is key to investigate their exact role in healthy ageing. To date, accelerometers using cut-off point models are most preferred for this, however, machine learning seems a highly promising future alternative. Hence, the current study compared between cut-off point and machine learning algorithms, for optimal quantification of sedentary behaviour and physical activity intensities in the elderly. Thus, in a heterogeneous sample of forty participants (aged ≥60 years, 50% female) energy expenditure during laboratory-based activities (ranging from sedentary behaviour through to moderate-to-vigorous physical activity) was estimated by indirect calorimetry, whilst wearing triaxial thigh-mounted accelerometers. Three cut-off point algorithms and a Random Forest machine learning model were developed and cross-validated using the collected data. Detailed analyses were performed to check algorithm robustness, and examine and benchmark both overall and participant-specific balanced accuracies. This revealed that the four models can at least be used to confidently monitor sedentary behaviour and moderate-to-vigorous physical activity. Nevertheless, the machine learning algorithm outperformed the cut-off point models by being robust for all individual's physiological and non-physiological characteristics and showing more performance of an acceptable level over the whole range of physical activity intensities. Therefore, we propose that Random Forest machine learning may be optimal for objective assessment of sedentary behaviour and physical activity in older adults using thigh-mounted triaxial accelerometry.

Can machine-learning improve cardiovascular risk prediction using routine clinical data?

PubMed Central

Kai, Joe; Garibaldi, Jonathan M.; Qureshi, Nadeem

2017-01-01

Background Current approaches to predict cardiovascular risk fail to identify many people who would benefit from preventive treatment, while others receive unnecessary intervention. Machine-learning offers opportunity to improve accuracy by exploiting complex interactions between risk factors. We assessed whether machine-learning can improve cardiovascular risk prediction. Methods Prospective cohort study using routine clinical data of 378,256 patients from UK family practices, free from cardiovascular disease at outset. Four machine-learning algorithms (random forest, logistic regression, gradient boosting machines, neural networks) were compared to an established algorithm (American College of Cardiology guidelines) to predict first cardiovascular event over 10-years. Predictive accuracy was assessed by area under the ‘receiver operating curve’ (AUC); and sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) to predict 7.5% cardiovascular risk (threshold for initiating statins). Findings 24,970 incident cardiovascular events (6.6%) occurred. Compared to the established risk prediction algorithm (AUC 0.728, 95% CI 0.723–0.735), machine-learning algorithms improved prediction: random forest +1.7% (AUC 0.745, 95% CI 0.739–0.750), logistic regression +3.2% (AUC 0.760, 95% CI 0.755–0.766), gradient boosting +3.3% (AUC 0.761, 95% CI 0.755–0.766), neural networks +3.6% (AUC 0.764, 95% CI 0.759–0.769). The highest achieving (neural networks) algorithm predicted 4,998/7,404 cases (sensitivity 67.5%, PPV 18.4%) and 53,458/75,585 non-cases (specificity 70.7%, NPV 95.7%), correctly predicting 355 (+7.6%) more patients who developed cardiovascular disease compared to the established algorithm. Conclusions Machine-learning significantly improves accuracy of cardiovascular risk prediction, increasing the number of patients identified who could benefit from preventive treatment, while avoiding unnecessary treatment of others. PMID:28376093

Can machine-learning improve cardiovascular risk prediction using routine clinical data?

PubMed

Weng, Stephen F; Reps, Jenna; Kai, Joe; Garibaldi, Jonathan M; Qureshi, Nadeem

2017-01-01

Current approaches to predict cardiovascular risk fail to identify many people who would benefit from preventive treatment, while others receive unnecessary intervention. Machine-learning offers opportunity to improve accuracy by exploiting complex interactions between risk factors. We assessed whether machine-learning can improve cardiovascular risk prediction. Prospective cohort study using routine clinical data of 378,256 patients from UK family practices, free from cardiovascular disease at outset. Four machine-learning algorithms (random forest, logistic regression, gradient boosting machines, neural networks) were compared to an established algorithm (American College of Cardiology guidelines) to predict first cardiovascular event over 10-years. Predictive accuracy was assessed by area under the 'receiver operating curve' (AUC); and sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) to predict 7.5% cardiovascular risk (threshold for initiating statins). 24,970 incident cardiovascular events (6.6%) occurred. Compared to the established risk prediction algorithm (AUC 0.728, 95% CI 0.723-0.735), machine-learning algorithms improved prediction: random forest +1.7% (AUC 0.745, 95% CI 0.739-0.750), logistic regression +3.2% (AUC 0.760, 95% CI 0.755-0.766), gradient boosting +3.3% (AUC 0.761, 95% CI 0.755-0.766), neural networks +3.6% (AUC 0.764, 95% CI 0.759-0.769). The highest achieving (neural networks) algorithm predicted 4,998/7,404 cases (sensitivity 67.5%, PPV 18.4%) and 53,458/75,585 non-cases (specificity 70.7%, NPV 95.7%), correctly predicting 355 (+7.6%) more patients who developed cardiovascular disease compared to the established algorithm. Machine-learning significantly improves accuracy of cardiovascular risk prediction, increasing the number of patients identified who could benefit from preventive treatment, while avoiding unnecessary treatment of others.

Machine learning methods for the classification of gliomas: Initial results using features extracted from MR spectroscopy.

PubMed

Ranjith, G; Parvathy, R; Vikas, V; Chandrasekharan, Kesavadas; Nair, Suresh

2015-04-01

With the advent of new imaging modalities, radiologists are faced with handling increasing volumes of data for diagnosis and treatment planning. The use of automated and intelligent systems is becoming essential in such a scenario. Machine learning, a branch of artificial intelligence, is increasingly being used in medical image analysis applications such as image segmentation, registration and computer-aided diagnosis and detection. Histopathological analysis is currently the gold standard for classification of brain tumors. The use of machine learning algorithms along with extraction of relevant features from magnetic resonance imaging (MRI) holds promise of replacing conventional invasive methods of tumor classification. The aim of the study is to classify gliomas into benign and malignant types using MRI data. Retrospective data from 28 patients who were diagnosed with glioma were used for the analysis. WHO Grade II (low-grade astrocytoma) was classified as benign while Grade III (anaplastic astrocytoma) and Grade IV (glioblastoma multiforme) were classified as malignant. Features were extracted from MR spectroscopy. The classification was done using four machine learning algorithms: multilayer perceptrons, support vector machine, random forest and locally weighted learning. Three of the four machine learning algorithms gave an area under ROC curve in excess of 0.80. Random forest gave the best performance in terms of AUC (0.911) while sensitivity was best for locally weighted learning (86.1%). The performance of different machine learning algorithms in the classification of gliomas is promising. An even better performance may be expected by integrating features extracted from other MR sequences. © The Author(s) 2015 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav.

Decision tree modeling using R.

PubMed

Zhang, Zhongheng

2016-08-01

In machine learning field, decision tree learner is powerful and easy to interpret. It employs recursive binary partitioning algorithm that splits the sample in partitioning variable with the strongest association with the response variable. The process continues until some stopping criteria are met. In the example I focus on conditional inference tree, which incorporates tree-structured regression models into conditional inference procedures. While growing a single tree is subject to small changes in the training data, random forests procedure is introduced to address this problem. The sources of diversity for random forests come from the random sampling and restricted set of input variables to be selected. Finally, I introduce R functions to perform model based recursive partitioning. This method incorporates recursive partitioning into conventional parametric model building.

Prediction of Baseflow Index of Catchments using Machine Learning Algorithms

NASA Astrophysics Data System (ADS)

Yadav, B.; Hatfield, K.

2017-12-01

We present the results of eight machine learning techniques for predicting the baseflow index (BFI) of ungauged basins using a surrogate of catchment scale climate and physiographic data. The tested algorithms include ordinary least squares, ridge regression, least absolute shrinkage and selection operator (lasso), elasticnet, support vector machine, gradient boosted regression trees, random forests, and extremely randomized trees. Our work seeks to identify the dominant controls of BFI that can be readily obtained from ancillary geospatial databases and remote sensing measurements, such that the developed techniques can be extended to ungauged catchments. More than 800 gauged catchments spanning the continental United States were selected to develop the general methodology. The BFI calculation was based on the baseflow separated from daily streamflow hydrograph using HYSEP filter. The surrogate catchment attributes were compiled from multiple sources including digital elevation model, soil, landuse, climate data, other publicly available ancillary and geospatial data. 80% catchments were used to train the ML algorithms, and the remaining 20% of the catchments were used as an independent test set to measure the generalization performance of fitted models. A k-fold cross-validation using exhaustive grid search was used to fit the hyperparameters of each model. Initial model development was based on 19 independent variables, but after variable selection and feature ranking, we generated revised sparse models of BFI prediction that are based on only six catchment attributes. These key predictive variables selected after the careful evaluation of bias-variance tradeoff include average catchment elevation, slope, fraction of sand, permeability, temperature, and precipitation. The most promising algorithms exceeding an accuracy score (r-square) of 0.7 on test data include support vector machine, gradient boosted regression trees, random forests, and extremely randomized trees. Considering both the accuracy and the computational complexity of these algorithms, we identify the extremely randomized trees as the best performing algorithm for BFI prediction in ungauged basins.

Predicting redox-sensitive contaminant concentrations in groundwater using random forest classification

NASA Astrophysics Data System (ADS)

Tesoriero, Anthony J.; Gronberg, Jo Ann; Juckem, Paul F.; Miller, Matthew P.; Austin, Brian P.

2017-08-01

Machine learning techniques were applied to a large (n > 10,000) compliance monitoring database to predict the occurrence of several redox-active constituents in groundwater across a large watershed. Specifically, random forest classification was used to determine the probabilities of detecting elevated concentrations of nitrate, iron, and arsenic in the Fox, Wolf, Peshtigo, and surrounding watersheds in northeastern Wisconsin. Random forest classification is well suited to describe the nonlinear relationships observed among several explanatory variables and the predicted probabilities of elevated concentrations of nitrate, iron, and arsenic. Maps of the probability of elevated nitrate, iron, and arsenic can be used to assess groundwater vulnerability and the vulnerability of streams to contaminants derived from groundwater. Processes responsible for elevated concentrations are elucidated using partial dependence plots. For example, an increase in the probability of elevated iron and arsenic occurred when well depths coincided with the glacial/bedrock interface, suggesting a bedrock source for these constituents. Furthermore, groundwater in contact with Ordovician bedrock has a higher likelihood of elevated iron concentrations, which supports the hypothesis that groundwater liberates iron from a sulfide-bearing secondary cement horizon of Ordovician age. Application of machine learning techniques to existing compliance monitoring data offers an opportunity to broadly assess aquifer and stream vulnerability at regional and national scales and to better understand geochemical processes responsible for observed conditions.

Predicting redox-sensitive contaminant concentrations in groundwater using random forest classification

USGS Publications Warehouse

Tesoriero, Anthony J.; Gronberg, Jo Ann M.; Juckem, Paul F.; Miller, Matthew P.; Austin, Brian P.

2017-01-01

Machine learning techniques were applied to a large (n > 10,000) compliance monitoring database to predict the occurrence of several redox-active constituents in groundwater across a large watershed. Specifically, random forest classification was used to determine the probabilities of detecting elevated concentrations of nitrate, iron, and arsenic in the Fox, Wolf, Peshtigo, and surrounding watersheds in northeastern Wisconsin. Random forest classification is well suited to describe the nonlinear relationships observed among several explanatory variables and the predicted probabilities of elevated concentrations of nitrate, iron, and arsenic. Maps of the probability of elevated nitrate, iron, and arsenic can be used to assess groundwater vulnerability and the vulnerability of streams to contaminants derived from groundwater. Processes responsible for elevated concentrations are elucidated using partial dependence plots. For example, an increase in the probability of elevated iron and arsenic occurred when well depths coincided with the glacial/bedrock interface, suggesting a bedrock source for these constituents. Furthermore, groundwater in contact with Ordovician bedrock has a higher likelihood of elevated iron concentrations, which supports the hypothesis that groundwater liberates iron from a sulfide-bearing secondary cement horizon of Ordovician age. Application of machine learning techniques to existing compliance monitoring data offers an opportunity to broadly assess aquifer and stream vulnerability at regional and national scales and to better understand geochemical processes responsible for observed conditions.

Integrating Archaeological Modeling in DoD Cultural Resource Compliance

DTIC Science & Technology

2012-10-26

Leo 2001 Random Forests. Machine Learning 45:5–32. Briuer, Frederick, Clifford Brown, Alan Gillespie, Fredrick Limp, Michael Trimble, and Len...glaciolacustrine clays on glacial lake plains Inceptisols Very-fine, mixed, active, nonacid, mesic Mollic Endoaquepts Low to none LoC Lowville silt

Combination of support vector machine, artificial neural network and random forest for improving the classification of convective and stratiform rain using spectral features of SEVIRI data

NASA Astrophysics Data System (ADS)

Lazri, Mourad; Ameur, Soltane

2018-05-01

A model combining three classifiers, namely Support vector machine, Artificial neural network and Random forest (SAR) is designed for improving the classification of convective and stratiform rain. This model (SAR model) has been trained and then tested on a datasets derived from MSG-SEVIRI (Meteosat Second Generation-Spinning Enhanced Visible and Infrared Imager). Well-classified, mid-classified and misclassified pixels are determined from the combination of three classifiers. Mid-classified and misclassified pixels that are considered unreliable pixels are reclassified by using a novel training of the developed scheme. In this novel training, only the input data corresponding to the pixels in question to are used. This whole process is repeated a second time and applied to mid-classified and misclassified pixels separately. Learning and validation of the developed scheme are realized against co-located data observed by ground radar. The developed scheme outperformed different classifiers used separately and reached 97.40% of overall accuracy of classification.

Discriminant forest classification method and system

DOEpatents

Chen, Barry Y.; Hanley, William G.; Lemmond, Tracy D.; Hiller, Lawrence J.; Knapp, David A.; Mugge, Marshall J.

2012-11-06

A hybrid machine learning methodology and system for classification that combines classical random forest (RF) methodology with discriminant analysis (DA) techniques to provide enhanced classification capability. A DA technique which uses feature measurements of an object to predict its class membership, such as linear discriminant analysis (LDA) or Andersen-Bahadur linear discriminant technique (AB), is used to split the data at each node in each of its classification trees to train and grow the trees and the forest. When training is finished, a set of n DA-based decision trees of a discriminant forest is produced for use in predicting the classification of new samples of unknown class.

Predicting membrane protein types using various decision tree classifiers based on various modes of general PseAAC for imbalanced datasets.

PubMed

Sankari, E Siva; Manimegalai, D

2017-12-21

Predicting membrane protein types is an important and challenging research area in bioinformatics and proteomics. Traditional biophysical methods are used to classify membrane protein types. Due to large exploration of uncharacterized protein sequences in databases, traditional methods are very time consuming, expensive and susceptible to errors. Hence, it is highly desirable to develop a robust, reliable, and efficient method to predict membrane protein types. Imbalanced datasets and large datasets are often handled well by decision tree classifiers. Since imbalanced datasets are taken, the performance of various decision tree classifiers such as Decision Tree (DT), Classification And Regression Tree (CART), C4.5, Random tree, REP (Reduced Error Pruning) tree, ensemble methods such as Adaboost, RUS (Random Under Sampling) boost, Rotation forest and Random forest are analysed. Among the various decision tree classifiers Random forest performs well in less time with good accuracy of 96.35%. Another inference is RUS boost decision tree classifier is able to classify one or two samples in the class with very less samples while the other classifiers such as DT, Adaboost, Rotation forest and Random forest are not sensitive for the classes with fewer samples. Also the performance of decision tree classifiers is compared with SVM (Support Vector Machine) and Naive Bayes classifier. Copyright © 2017 Elsevier Ltd. All rights reserved.

The Efficiency of Random Forest Method for Shoreline Extraction from LANDSAT-8 and GOKTURK-2 Imageries

NASA Astrophysics Data System (ADS)

Bayram, B.; Erdem, F.; Akpinar, B.; Ince, A. K.; Bozkurt, S.; Catal Reis, H.; Seker, D. Z.

2017-11-01

Coastal monitoring plays a vital role in environmental planning and hazard management related issues. Since shorelines are fundamental data for environment management, disaster management, coastal erosion studies, modelling of sediment transport and coastal morphodynamics, various techniques have been developed to extract shorelines. Random Forest is one of these techniques which is used in this study for shoreline extraction.. This algorithm is a machine learning method based on decision trees. Decision trees analyse classes of training data creates rules for classification. In this study, Terkos region has been chosen for the proposed method within the scope of "TUBITAK Project (Project No: 115Y718) titled "Integration of Unmanned Aerial Vehicles for Sustainable Coastal Zone Monitoring Model - Three-Dimensional Automatic Coastline Extraction and Analysis: Istanbul-Terkos Example". Random Forest algorithm has been implemented to extract the shoreline of the Black Sea where near the lake from LANDSAT-8 and GOKTURK-2 satellite imageries taken in 2015. The MATLAB environment was used for classification. To obtain land and water-body classes, the Random Forest method has been applied to NIR bands of LANDSAT-8 (5th band) and GOKTURK-2 (4th band) imageries. Each image has been digitized manually and shorelines obtained for accuracy assessment. According to accuracy assessment results, Random Forest method is efficient for both medium and high resolution images for shoreline extraction studies.

Assessment of various supervised learning algorithms using different performance metrics

NASA Astrophysics Data System (ADS)

Susheel Kumar, S. M.; Laxkar, Deepak; Adhikari, Sourav; Vijayarajan, V.

2017-11-01

Our work brings out comparison based on the performance of supervised machine learning algorithms on a binary classification task. The supervised machine learning algorithms which are taken into consideration in the following work are namely Support Vector Machine(SVM), Decision Tree(DT), K Nearest Neighbour (KNN), Naïve Bayes(NB) and Random Forest(RF). This paper mostly focuses on comparing the performance of above mentioned algorithms on one binary classification task by analysing the Metrics such as Accuracy, F-Measure, G-Measure, Precision, Misclassification Rate, False Positive Rate, True Positive Rate, Specificity, Prevalence.

Does Sentinel multi sensor data offer synergy in Improving Accuracy of Aboveground Biomass Estimate of Dense Tropical Forest? - Utility of Decision Tree Based Machine Learning Algorithms

NASA Astrophysics Data System (ADS)

Ghosh, S. M.; Behera, M. D.

2017-12-01

Forest aboveground biomass (AGB) is an important factor for preparation of global policy making decisions to tackle the impact of climate change. Several previous studies has concluded that remote sensing methods are more suitable for estimating forest biomass on regional scale. Among all available remote sensing data and methods, Synthetic Aperture Radar (SAR) data in combination with decision tree based machine learning algorithms has shown better promise in estimating higher biomass values. There aren't many studies done for biomass estimation of dense Indian tropical forests with high biomass density. In this study aboveground biomass was estimated for two major tree species, Sal (Shorea robusta) and Teak (Tectona grandis), of Katerniaghat Wildlife Sanctuary, a tropical forest situated in northern India. Biomass was estimated by combining C-band SAR data from Sentinel-1A satellite, vegetation indices produced using Sentinel-2A data and ground inventory plots. Along with SAR backscatter value, SAR texture images were also used as input as earlier studies had found that image texture has a correlation with vegetation biomass. Decision tree based nonlinear machine learning algorithms were used in place of parametric regression models for establishing relationship between fields measured values and remotely sensed parameters. Using random forest model with a combination of vegetation indices with SAR backscatter as predictor variables shows best result for Sal forest, with a coefficient of determination value of 0.71 and a RMSE value of 105.027 t/ha. In teak forest also best result can be found in the same combination but for stochastic gradient boosted model with a coefficient of determination value of 0.6 and a RMSE value of 79.45 t/ha. These results are mostly better than the results of other studies done for similar kind of forests. This study shows that Sentinel series satellite data has exceptional capabilities in estimating dense forest AGB and machine learning algorithms are better means to do so than parametric regression models.

Ensemble Feature Learning of Genomic Data Using Support Vector Machine

PubMed Central

Anaissi, Ali; Goyal, Madhu; Catchpoole, Daniel R.; Braytee, Ali; Kennedy, Paul J.

2016-01-01

The identification of a subset of genes having the ability to capture the necessary information to distinguish classes of patients is crucial in bioinformatics applications. Ensemble and bagging methods have been shown to work effectively in the process of gene selection and classification. Testament to that is random forest which combines random decision trees with bagging to improve overall feature selection and classification accuracy. Surprisingly, the adoption of these methods in support vector machines has only recently received attention but mostly on classification not gene selection. This paper introduces an ensemble SVM-Recursive Feature Elimination (ESVM-RFE) for gene selection that follows the concepts of ensemble and bagging used in random forest but adopts the backward elimination strategy which is the rationale of RFE algorithm. The rationale behind this is, building ensemble SVM models using randomly drawn bootstrap samples from the training set, will produce different feature rankings which will be subsequently aggregated as one feature ranking. As a result, the decision for elimination of features is based upon the ranking of multiple SVM models instead of choosing one particular model. Moreover, this approach will address the problem of imbalanced datasets by constructing a nearly balanced bootstrap sample. Our experiments show that ESVM-RFE for gene selection substantially increased the classification performance on five microarray datasets compared to state-of-the-art methods. Experiments on the childhood leukaemia dataset show that an average 9% better accuracy is achieved by ESVM-RFE over SVM-RFE, and 5% over random forest based approach. The selected genes by the ESVM-RFE algorithm were further explored with Singular Value Decomposition (SVD) which reveals significant clusters with the selected data. PMID:27304923

FOCIS: A forest classification and inventory system using LANDSAT and digital terrain data

NASA Technical Reports Server (NTRS)

Strahler, A. H.; Franklin, J.; Woodcook, C. E.; Logan, T. L.

1981-01-01

Accurate, cost-effective stratification of forest vegetation and timber inventory is the primary goal of a Forest Classification and Inventory System (FOCIS). Conventional timber stratification using photointerpretation can be time-consuming, costly, and inconsistent from analyst to analyst. FOCIS was designed to overcome these problems by using machine processing techniques to extract and process tonal, textural, and terrain information from registered LANDSAT multispectral and digital terrain data. Comparison of samples from timber strata identified by conventional procedures showed that both have about the same potential to reduce the variance of timber volume estimates over simple random sampling.

Machine Learning Methods for Production Cases Analysis

NASA Astrophysics Data System (ADS)

Mokrova, Nataliya V.; Mokrov, Alexander M.; Safonova, Alexandra V.; Vishnyakov, Igor V.

2018-03-01

Approach to analysis of events occurring during the production process were proposed. Described machine learning system is able to solve classification tasks related to production control and hazard identification at an early stage. Descriptors of the internal production network data were used for training and testing of applied models. k-Nearest Neighbors and Random forest methods were used to illustrate and analyze proposed solution. The quality of the developed classifiers was estimated using standard statistical metrics, such as precision, recall and accuracy.

Combining macula clinical signs and patient characteristics for age-related macular degeneration diagnosis: a machine learning approach.

PubMed

Fraccaro, Paolo; Nicolo, Massimo; Bonetto, Monica; Giacomini, Mauro; Weller, Peter; Traverso, Carlo Enrico; Prosperi, Mattia; OSullivan, Dympna

2015-01-27

To investigate machine learning methods, ranging from simpler interpretable techniques to complex (non-linear) "black-box" approaches, for automated diagnosis of Age-related Macular Degeneration (AMD). Data from healthy subjects and patients diagnosed with AMD or other retinal diseases were collected during routine visits via an Electronic Health Record (EHR) system. Patients' attributes included demographics and, for each eye, presence/absence of major AMD-related clinical signs (soft drusen, retinal pigment epitelium, defects/pigment mottling, depigmentation area, subretinal haemorrhage, subretinal fluid, macula thickness, macular scar, subretinal fibrosis). Interpretable techniques known as white box methods including logistic regression and decision trees as well as less interpreitable techniques known as black box methods, such as support vector machines (SVM), random forests and AdaBoost, were used to develop models (trained and validated on unseen data) to diagnose AMD. The gold standard was confirmed diagnosis of AMD by physicians. Sensitivity, specificity and area under the receiver operating characteristic (AUC) were used to assess performance. Study population included 487 patients (912 eyes). In terms of AUC, random forests, logistic regression and adaboost showed a mean performance of (0.92), followed by SVM and decision trees (0.90). All machine learning models identified soft drusen and age as the most discriminating variables in clinicians' decision pathways to diagnose AMD. Both black-box and white box methods performed well in identifying diagnoses of AMD and their decision pathways. Machine learning models developed through the proposed approach, relying on clinical signs identified by retinal specialists, could be embedded into EHR to provide physicians with real time (interpretable) support.

Random forest models to predict aqueous solubility.

PubMed

Palmer, David S; O'Boyle, Noel M; Glen, Robert C; Mitchell, John B O

2007-01-01

Random Forest regression (RF), Partial-Least-Squares (PLS) regression, Support Vector Machines (SVM), and Artificial Neural Networks (ANN) were used to develop QSPR models for the prediction of aqueous solubility, based on experimental data for 988 organic molecules. The Random Forest regression model predicted aqueous solubility more accurately than those created by PLS, SVM, and ANN and offered methods for automatic descriptor selection, an assessment of descriptor importance, and an in-parallel measure of predictive ability, all of which serve to recommend its use. The prediction of log molar solubility for an external test set of 330 molecules that are solid at 25 degrees C gave an r2 = 0.89 and RMSE = 0.69 log S units. For a standard data set selected from the literature, the model performed well with respect to other documented methods. Finally, the diversity of the training and test sets are compared to the chemical space occupied by molecules in the MDL drug data report, on the basis of molecular descriptors selected by the regression analysis.

Integration of spectral, spatial and morphometric data into lithological mapping: A comparison of different Machine Learning Algorithms in the Kurdistan Region, NE Iraq

NASA Astrophysics Data System (ADS)

Othman, Arsalan A.; Gloaguen, Richard

2017-09-01

Lithological mapping in mountainous regions is often impeded by limited accessibility due to relief. This study aims to evaluate (1) the performance of different supervised classification approaches using remote sensing data and (2) the use of additional information such as geomorphology. We exemplify the methodology in the Bardi-Zard area in NE Iraq, a part of the Zagros Fold - Thrust Belt, known for its chromite deposits. We highlighted the improvement of remote sensing geological classification by integrating geomorphic features and spatial information in the classification scheme. We performed a Maximum Likelihood (ML) classification method besides two Machine Learning Algorithms (MLA): Support Vector Machine (SVM) and Random Forest (RF) to allow the joint use of geomorphic features, Band Ratio (BR), Principal Component Analysis (PCA), spatial information (spatial coordinates) and multispectral data of the Advanced Space-borne Thermal Emission and Reflection radiometer (ASTER) satellite. The RF algorithm showed reliable results and discriminated serpentinite, talus and terrace deposits, red argillites with conglomerates and limestone, limy conglomerates and limestone conglomerates, tuffites interbedded with basic lavas, limestone and Metamorphosed limestone and reddish green shales. The best overall accuracy (∼80%) was achieved by Random Forest (RF) algorithms in the majority of the sixteen tested combination datasets.

Predicting human liver microsomal stability with machine learning techniques.

PubMed

Sakiyama, Yojiro; Yuki, Hitomi; Moriya, Takashi; Hattori, Kazunari; Suzuki, Misaki; Shimada, Kaoru; Honma, Teruki

2008-02-01

To ensure a continuing pipeline in pharmaceutical research, lead candidates must possess appropriate metabolic stability in the drug discovery process. In vitro ADMET (absorption, distribution, metabolism, elimination, and toxicity) screening provides us with useful information regarding the metabolic stability of compounds. However, before the synthesis stage, an efficient process is required in order to deal with the vast quantity of data from large compound libraries and high-throughput screening. Here we have derived a relationship between the chemical structure and its metabolic stability for a data set of in-house compounds by means of various in silico machine learning such as random forest, support vector machine (SVM), logistic regression, and recursive partitioning. For model building, 1952 proprietary compounds comprising two classes (stable/unstable) were used with 193 descriptors calculated by Molecular Operating Environment. The results using test compounds have demonstrated that all classifiers yielded satisfactory results (accuracy > 0.8, sensitivity > 0.9, specificity > 0.6, and precision > 0.8). Above all, classification by random forest as well as SVM yielded kappa values of approximately 0.7 in an independent validation set, slightly higher than other classification tools. These results suggest that nonlinear/ensemble-based classification methods might prove useful in the area of in silico ADME modeling.

Mortality risk score prediction in an elderly population using machine learning.

PubMed

Rose, Sherri

2013-03-01

Standard practice for prediction often relies on parametric regression methods. Interesting new methods from the machine learning literature have been introduced in epidemiologic studies, such as random forest and neural networks. However, a priori, an investigator will not know which algorithm to select and may wish to try several. Here I apply the super learner, an ensembling machine learning approach that combines multiple algorithms into a single algorithm and returns a prediction function with the best cross-validated mean squared error. Super learning is a generalization of stacking methods. I used super learning in the Study of Physical Performance and Age-Related Changes in Sonomans (SPPARCS) to predict death among 2,066 residents of Sonoma, California, aged 54 years or more during the period 1993-1999. The super learner for predicting death (risk score) improved upon all single algorithms in the collection of algorithms, although its performance was similar to that of several algorithms. Super learner outperformed the worst algorithm (neural networks) by 44% with respect to estimated cross-validated mean squared error and had an R2 value of 0.201. The improvement of super learner over random forest with respect to R2 was approximately 2-fold. Alternatives for risk score prediction include the super learner, which can provide improved performance.

A machine learning model with human cognitive biases capable of learning from small and biased datasets.

PubMed

Taniguchi, Hidetaka; Sato, Hiroshi; Shirakawa, Tomohiro

2018-05-09

Human learners can generalize a new concept from a small number of samples. In contrast, conventional machine learning methods require large amounts of data to address the same types of problems. Humans have cognitive biases that promote fast learning. Here, we developed a method to reduce the gap between human beings and machines in this type of inference by utilizing cognitive biases. We implemented a human cognitive model into machine learning algorithms and compared their performance with the currently most popular methods, naïve Bayes, support vector machine, neural networks, logistic regression and random forests. We focused on the task of spam classification, which has been studied for a long time in the field of machine learning and often requires a large amount of data to obtain high accuracy. Our models achieved superior performance with small and biased samples in comparison with other representative machine learning methods.

Modeling Verdict Outcomes Using Social Network Measures: The Watergate and Caviar Network Cases.

PubMed

Masías, Víctor Hugo; Valle, Mauricio; Morselli, Carlo; Crespo, Fernando; Vargas, Augusto; Laengle, Sigifredo

2016-01-01

Modelling criminal trial verdict outcomes using social network measures is an emerging research area in quantitative criminology. Few studies have yet analyzed which of these measures are the most important for verdict modelling or which data classification techniques perform best for this application. To compare the performance of different techniques in classifying members of a criminal network, this article applies three different machine learning classifiers-Logistic Regression, Naïve Bayes and Random Forest-with a range of social network measures and the necessary databases to model the verdicts in two real-world cases: the U.S. Watergate Conspiracy of the 1970's and the now-defunct Canada-based international drug trafficking ring known as the Caviar Network. In both cases it was found that the Random Forest classifier did better than either Logistic Regression or Naïve Bayes, and its superior performance was statistically significant. This being so, Random Forest was used not only for classification but also to assess the importance of the measures. For the Watergate case, the most important one proved to be betweenness centrality while for the Caviar Network, it was the effective size of the network. These results are significant because they show that an approach combining machine learning with social network analysis not only can generate accurate classification models but also helps quantify the importance social network variables in modelling verdict outcomes. We conclude our analysis with a discussion and some suggestions for future work in verdict modelling using social network measures.

Comparison of machine-learning methods for above-ground biomass estimation based on Landsat imagery

NASA Astrophysics Data System (ADS)

Wu, Chaofan; Shen, Huanhuan; Shen, Aihua; Deng, Jinsong; Gan, Muye; Zhu, Jinxia; Xu, Hongwei; Wang, Ke

2016-07-01

Biomass is one significant biophysical parameter of a forest ecosystem, and accurate biomass estimation on the regional scale provides important information for carbon-cycle investigation and sustainable forest management. In this study, Landsat satellite imagery data combined with field-based measurements were integrated through comparisons of five regression approaches [stepwise linear regression, K-nearest neighbor, support vector regression, random forest (RF), and stochastic gradient boosting] with two different candidate variable strategies to implement the optimal spatial above-ground biomass (AGB) estimation. The results suggested that RF algorithm exhibited the best performance by 10-fold cross-validation with respect to R2 (0.63) and root-mean-square error (26.44 ton/ha). Consequently, the map of estimated AGB was generated with a mean value of 89.34 ton/ha in northwestern Zhejiang Province, China, with a similar pattern to the distribution mode of local forest species. This research indicates that machine-learning approaches associated with Landsat imagery provide an economical way for biomass estimation. Moreover, ensemble methods using all candidate variables, especially for Landsat images, provide an alternative for regional biomass simulation.

TH-CD-206-05: Machine-Learning Based Segmentation of Organs at Risks for Head and Neck Radiotherapy Planning

DOE Office of Scientific and Technical Information (OSTI.GOV)

Ibragimov, B; Pernus, F; Strojan, P

Purpose: Accurate and efficient delineation of tumor target and organs-at-risks is essential for the success of radiotherapy. In reality, despite of decades of intense research efforts, auto-segmentation has not yet become clinical practice. In this study, we present, for the first time, a deep learning-based classification algorithm for autonomous segmentation in head and neck (HaN) treatment planning. Methods: Fifteen HN datasets of CT, MR and PET images with manual annotation of organs-at-risk (OARs) including spinal cord, brainstem, optic nerves, chiasm, eyes, mandible, tongue, parotid glands were collected and saved in a library of plans. We also have ten super-resolution MRmore » images of the tongue area, where the genioglossus and inferior longitudinalis tongue muscles are defined as organs of interest. We applied the concepts of random forest- and deep learning-based object classification for automated image annotation with the aim of using machine learning to facilitate head and neck radiotherapy planning process. In this new paradigm of segmentation, random forests were used for landmark-assisted segmentation of super-resolution MR images. Alternatively to auto-segmentation with random forest-based landmark detection, deep convolutional neural networks were developed for voxel-wise segmentation of OARs in single and multi-modal images. The network consisted of three pairs of convolution and pooing layer, one RuLU layer and a softmax layer. Results: We present a comprehensive study on using machine learning concepts for auto-segmentation of OARs and tongue muscles for the HaN radiotherapy planning. An accuracy of 81.8% in terms of Dice coefficient was achieved for segmentation of genioglossus and inferior longitudinalis tongue muscles. Preliminary results of OARs regimentation also indicate that deep-learning afforded an unprecedented opportunities to improve the accuracy and robustness of radiotherapy planning. Conclusion: A novel machine learning framework has been developed for image annotation and structure segmentation. Our results indicate the great potential of deep learning in radiotherapy treatment planning.« less

Cardiovascular Event Prediction by Machine Learning: The Multi-Ethnic Study of Atherosclerosis.

PubMed

Ambale-Venkatesh, Bharath; Yang, Xiaoying; Wu, Colin O; Liu, Kiang; Hundley, W Gregory; McClelland, Robyn; Gomes, Antoinette S; Folsom, Aaron R; Shea, Steven; Guallar, Eliseo; Bluemke, David A; Lima, João A C

2017-10-13

Machine learning may be useful to characterize cardiovascular risk, predict outcomes, and identify biomarkers in population studies. To test the ability of random survival forests, a machine learning technique, to predict 6 cardiovascular outcomes in comparison to standard cardiovascular risk scores. We included participants from the MESA (Multi-Ethnic Study of Atherosclerosis). Baseline measurements were used to predict cardiovascular outcomes over 12 years of follow-up. MESA was designed to study progression of subclinical disease to cardiovascular events where participants were initially free of cardiovascular disease. All 6814 participants from MESA, aged 45 to 84 years, from 4 ethnicities, and 6 centers across the United States were included. Seven-hundred thirty-five variables from imaging and noninvasive tests, questionnaires, and biomarker panels were obtained. We used the random survival forests technique to identify the top-20 predictors of each outcome. Imaging, electrocardiography, and serum biomarkers featured heavily on the top-20 lists as opposed to traditional cardiovascular risk factors. Age was the most important predictor for all-cause mortality. Fasting glucose levels and carotid ultrasonography measures were important predictors of stroke. Coronary Artery Calcium score was the most important predictor of coronary heart disease and all atherosclerotic cardiovascular disease combined outcomes. Left ventricular structure and function and cardiac troponin-T were among the top predictors for incident heart failure. Creatinine, age, and ankle-brachial index were among the top predictors of atrial fibrillation. TNF-α (tissue necrosis factor-α) and IL (interleukin)-2 soluble receptors and NT-proBNP (N-Terminal Pro-B-Type Natriuretic Peptide) levels were important across all outcomes. The random survival forests technique performed better than established risk scores with increased prediction accuracy (decreased Brier score by 10%-25%). Machine learning in conjunction with deep phenotyping improves prediction accuracy in cardiovascular event prediction in an initially asymptomatic population. These methods may lead to greater insights on subclinical disease markers without apriori assumptions of causality. URL: http://www.clinicaltrials.gov. Unique identifier: NCT00005487. © 2017 American Heart Association, Inc.

Disruption prediction investigations using Machine Learning tools on DIII-D and Alcator C-Mod

DOE Office of Scientific and Technical Information (OSTI.GOV)

Rea, C.; Granetz, R. S.; Montes, K.

Using data-driven methodology, we exploit the time series of relevant plasma parameters for a large set of disrupted and non-disrupted discharges to develop a classification algorithm for detecting disruptive phases in shots that eventually disrupt. Comparing the same methodology on different devices is crucial in order to have information on the portability of the developed algorithm and the possible extrapolation to ITER. Therefore, we use data from two very different tokamaks, DIII-D and Alcator C-Mod. We then focus on a subset of disruption predictors, most of which are dimensionless and/or machine-independent parameters, coming from both plasma diagnostics and equilibrium reconstructions,more » such as the normalized plasma internal inductance ℓ and the n = 1 mode amplitude normalized to the toroidal magnetic field. Using such dimensionless indicators facilitates a more direct comparison between DIII-D and C-Mod. We then choose a shallow Machine Learning technique, called Random Forests, to explore the databases available for the two devices. We show results from the classification task, where we introduce a time dependency through the definition of class labels on the basis of the elapsed time before the disruption (i.e. ‘far from a disruption’ and ‘close to a disruption’). The performances of the different Random Forest classifiers are discussed in terms of several metrics, by showing the number of successfully detected samples, as well as the misclassifications. The overall model accuracies are above 97% when identifying a ‘far from disruption’ and a ‘disruptive’ phase for disrupted discharges. Nevertheless, the Forests are intrinsically different in their capability of predicting a disruptive behavior, with C-Mod predictions comparable to random guesses. Indeed, we show that C-Mod recall index, i.e. the sensitivity to a disruptive behavior, is as low as 0.47, while DIII-D recall is ~0.72. The portability of the developed algorithm is also tested across the two devices, by using DIII-D data for training the forests and C-Mod for testing and vice versa.« less

Disruption prediction investigations using Machine Learning tools on DIII-D and Alcator C-Mod

DOE PAGES

Rea, C.; Granetz, R. S.; Montes, K.; ...

2018-06-18

Using data-driven methodology, we exploit the time series of relevant plasma parameters for a large set of disrupted and non-disrupted discharges to develop a classification algorithm for detecting disruptive phases in shots that eventually disrupt. Comparing the same methodology on different devices is crucial in order to have information on the portability of the developed algorithm and the possible extrapolation to ITER. Therefore, we use data from two very different tokamaks, DIII-D and Alcator C-Mod. We then focus on a subset of disruption predictors, most of which are dimensionless and/or machine-independent parameters, coming from both plasma diagnostics and equilibrium reconstructions,more » such as the normalized plasma internal inductance ℓ and the n = 1 mode amplitude normalized to the toroidal magnetic field. Using such dimensionless indicators facilitates a more direct comparison between DIII-D and C-Mod. We then choose a shallow Machine Learning technique, called Random Forests, to explore the databases available for the two devices. We show results from the classification task, where we introduce a time dependency through the definition of class labels on the basis of the elapsed time before the disruption (i.e. ‘far from a disruption’ and ‘close to a disruption’). The performances of the different Random Forest classifiers are discussed in terms of several metrics, by showing the number of successfully detected samples, as well as the misclassifications. The overall model accuracies are above 97% when identifying a ‘far from disruption’ and a ‘disruptive’ phase for disrupted discharges. Nevertheless, the Forests are intrinsically different in their capability of predicting a disruptive behavior, with C-Mod predictions comparable to random guesses. Indeed, we show that C-Mod recall index, i.e. the sensitivity to a disruptive behavior, is as low as 0.47, while DIII-D recall is ~0.72. The portability of the developed algorithm is also tested across the two devices, by using DIII-D data for training the forests and C-Mod for testing and vice versa.« less

Random forest feature selection, fusion and ensemble strategy: Combining multiple morphological MRI measures to discriminate among healhy elderly, MCI, cMCI and alzheimer's disease patients: From the alzheimer's disease neuroimaging initiative (ADNI) database.

PubMed

Dimitriadis, S I; Liparas, Dimitris; Tsolaki, Magda N

2018-05-15

In the era of computer-assisted diagnostic tools for various brain diseases, Alzheimer's disease (AD) covers a large percentage of neuroimaging research, with the main scope being its use in daily practice. However, there has been no study attempting to simultaneously discriminate among Healthy Controls (HC), early mild cognitive impairment (MCI), late MCI (cMCI) and stable AD, using features derived from a single modality, namely MRI. Based on preprocessed MRI images from the organizers of a neuroimaging challenge, 3 we attempted to quantify the prediction accuracy of multiple morphological MRI features to simultaneously discriminate among HC, MCI, cMCI and AD. We explored the efficacy of a novel scheme that includes multiple feature selections via Random Forest from subsets of the whole set of features (e.g. whole set, left/right hemisphere etc.), Random Forest classification using a fusion approach and ensemble classification via majority voting. From the ADNI database, 60 HC, 60 MCI, 60 cMCI and 60 CE were used as a training set with known labels. An extra dataset of 160 subjects (HC: 40, MCI: 40, cMCI: 40 and AD: 40) was used as an external blind validation dataset to evaluate the proposed machine learning scheme. In the second blind dataset, we succeeded in a four-class classification of 61.9% by combining MRI-based features with a Random Forest-based Ensemble Strategy. We achieved the best classification accuracy of all teams that participated in this neuroimaging competition. The results demonstrate the effectiveness of the proposed scheme to simultaneously discriminate among four groups using morphological MRI features for the very first time in the literature. Hence, the proposed machine learning scheme can be used to define single and multi-modal biomarkers for AD. Copyright © 2017 Elsevier B.V. All rights reserved.

A random forest learning assisted "divide and conquer" approach for peptide conformation search.

PubMed

Chen, Xin; Yang, Bing; Lin, Zijing

2018-06-11

Computational determination of peptide conformations is challenging as it is a problem of finding minima in a high-dimensional space. The "divide and conquer" approach is promising for reliably reducing the search space size. A random forest learning model is proposed here to expand the scope of applicability of the "divide and conquer" approach. A random forest classification algorithm is used to characterize the distributions of the backbone φ-ψ units ("words"). A random forest supervised learning model is developed to analyze the combinations of the φ-ψ units ("grammar"). It is found that amino acid residues may be grouped as equivalent "words", while the φ-ψ combinations in low-energy peptide conformations follow a distinct "grammar". The finding of equivalent words empowers the "divide and conquer" method with the flexibility of fragment substitution. The learnt grammar is used to improve the efficiency of the "divide and conquer" method by removing unfavorable φ-ψ combinations without the need of dedicated human effort. The machine learning assisted search method is illustrated by efficiently searching the conformations of GGG/AAA/GGGG/AAAA/GGGGG through assembling the structures of GFG/GFGG. Moreover, the computational cost of the new method is shown to increase rather slowly with the peptide length.

Random Forest as a Predictive Analytics Alternative to Regression in Institutional Research

ERIC Educational Resources Information Center

He, Lingjun; Levine, Richard A.; Fan, Juanjuan; Beemer, Joshua; Stronach, Jeanne

2018-01-01

In institutional research, modern data mining approaches are seldom considered to address predictive analytics problems. The goal of this paper is to highlight the advantages of tree-based machine learning algorithms over classic (logistic) regression methods for data-informed decision making in higher education problems, and stress the success of…

Per-field crop classification in irrigated agricultural regions in middle Asia using random forest and support vector machine ensemble

NASA Astrophysics Data System (ADS)

Löw, Fabian; Schorcht, Gunther; Michel, Ulrich; Dech, Stefan; Conrad, Christopher

2012-10-01

Accurate crop identification and crop area estimation are important for studies on irrigated agricultural systems, yield and water demand modeling, and agrarian policy development. In this study a novel combination of Random Forest (RF) and Support Vector Machine (SVM) classifiers is presented that (i) enhances crop classification accuracy and (ii) provides spatial information on map uncertainty. The methodology was implemented over four distinct irrigated sites in Middle Asia using RapidEye time series data. The RF feature importance statistics was used as feature-selection strategy for the SVM to assess possible negative effects on classification accuracy caused by an oversized feature space. The results of the individual RF and SVM classifications were combined with rules based on posterior classification probability and estimates of classification probability entropy. SVM classification performance was increased by feature selection through RF. Further experimental results indicate that the hybrid classifier improves overall classification accuracy in comparison to the single classifiers as well as useŕs and produceŕs accuracy.

Modelling Biophysical Parameters of Maize Using Landsat 8 Time Series

NASA Astrophysics Data System (ADS)

Dahms, Thorsten; Seissiger, Sylvia; Conrad, Christopher; Borg, Erik

2016-06-01

Open and free access to multi-frequent high-resolution data (e.g. Sentinel - 2) will fortify agricultural applications based on satellite data. The temporal and spatial resolution of these remote sensing datasets directly affects the applicability of remote sensing methods, for instance a robust retrieving of biophysical parameters over the entire growing season with very high geometric resolution. In this study we use machine learning methods to predict biophysical parameters, namely the fraction of absorbed photosynthetic radiation (FPAR), the leaf area index (LAI) and the chlorophyll content, from high resolution remote sensing. 30 Landsat 8 OLI scenes were available in our study region in Mecklenburg-Western Pomerania, Germany. In-situ data were weekly to bi-weekly collected on 18 maize plots throughout the summer season 2015. The study aims at an optimized prediction of biophysical parameters and the identification of the best explaining spectral bands and vegetation indices. For this purpose, we used the entire in-situ dataset from 24.03.2015 to 15.10.2015. Random forest and conditional inference forests were used because of their explicit strong exploratory and predictive character. Variable importance measures allowed for analysing the relation between the biophysical parameters with respect to the spectral response, and the performance of the two approaches over the plant stock evolvement. Classical random forest regression outreached the performance of conditional inference forests, in particular when modelling the biophysical parameters over the entire growing period. For example, modelling biophysical parameters of maize for the entire vegetation period using random forests yielded: FPAR: R² = 0.85; RMSE = 0.11; LAI: R² = 0.64; RMSE = 0.9 and chlorophyll content (SPAD): R² = 0.80; RMSE=4.9. Our results demonstrate the great potential in using machine-learning methods for the interpretation of long-term multi-frequent remote sensing datasets to model biophysical parameters.

Recent advances in environmental data mining

NASA Astrophysics Data System (ADS)

Leuenberger, Michael; Kanevski, Mikhail

2016-04-01

Due to the large amount and complexity of data available nowadays in geo- and environmental sciences, we face the need to develop and incorporate more robust and efficient methods for their analysis, modelling and visualization. An important part of these developments deals with an elaboration and application of a contemporary and coherent methodology following the process from data collection to the justification and communication of the results. Recent fundamental progress in machine learning (ML) can considerably contribute to the development of the emerging field - environmental data science. The present research highlights and investigates the different issues that can occur when dealing with environmental data mining using cutting-edge machine learning algorithms. In particular, the main attention is paid to the description of the self-consistent methodology and two efficient algorithms - Random Forest (RF, Breiman, 2001) and Extreme Learning Machines (ELM, Huang et al., 2006), which recently gained a great popularity. Despite the fact that they are based on two different concepts, i.e. decision trees vs artificial neural networks, they both propose promising results for complex, high dimensional and non-linear data modelling. In addition, the study discusses several important issues of data driven modelling, including feature selection and uncertainties. The approach considered is accompanied by simulated and real data case studies from renewable resources assessment and natural hazards tasks. In conclusion, the current challenges and future developments in statistical environmental data learning are discussed. References - Breiman, L., 2001. Random Forests. Machine Learning 45 (1), 5-32. - Huang, G.-B., Zhu, Q.-Y., Siew, C.-K., 2006. Extreme learning machine: theory and applications. Neurocomputing 70 (1-3), 489-501. - Kanevski, M., Pozdnoukhov, A., Timonin, V., 2009. Machine Learning for Spatial Environmental Data. EPFL Press; Lausanne, Switzerland, p.392. - Leuenberger, M., Kanevski, M., 2015. Extreme Learning Machines for spatial environmental data. Computers and Geosciences 85, 64-73.

Fiber tractography using machine learning.

PubMed

Neher, Peter F; Côté, Marc-Alexandre; Houde, Jean-Christophe; Descoteaux, Maxime; Maier-Hein, Klaus H

2017-09-01

We present a fiber tractography approach based on a random forest classification and voting process, guiding each step of the streamline progression by directly processing raw diffusion-weighted signal intensities. For comparison to the state-of-the-art, i.e. tractography pipelines that rely on mathematical modeling, we performed a quantitative and qualitative evaluation with multiple phantom and in vivo experiments, including a comparison to the 96 submissions of the ISMRM tractography challenge 2015. The results demonstrate the vast potential of machine learning for fiber tractography. Copyright © 2017 Elsevier Inc. All rights reserved.

Epidermis area detection for immunofluorescence microscopy

NASA Astrophysics Data System (ADS)

Dovganich, Andrey; Krylov, Andrey; Nasonov, Andrey; Makhneva, Natalia

2018-04-01

We propose a novel image segmentation method for immunofluorescence microscopy images of skin tissue for the diagnosis of various skin diseases. The segmentation is based on machine learning algorithms. The feature vector is filled by three groups of features: statistical features, Laws' texture energy measures and local binary patterns. The images are preprocessed for better learning. Different machine learning algorithms have been used and the best results have been obtained with random forest algorithm. We use the proposed method to detect the epidermis region as a part of pemphigus diagnosis system.

High-Risk Breast Lesions: A Machine Learning Model to Predict Pathologic Upgrade and Reduce Unnecessary Surgical Excision.

PubMed

Bahl, Manisha; Barzilay, Regina; Yedidia, Adam B; Locascio, Nicholas J; Yu, Lili; Lehman, Constance D

2018-03-01

Purpose To develop a machine learning model that allows high-risk breast lesions (HRLs) diagnosed with image-guided needle biopsy that require surgical excision to be distinguished from HRLs that are at low risk for upgrade to cancer at surgery and thus could be surveilled. Materials and Methods Consecutive patients with biopsy-proven HRLs who underwent surgery or at least 2 years of imaging follow-up from June 2006 to April 2015 were identified. A random forest machine learning model was developed to identify HRLs at low risk for upgrade to cancer. Traditional features such as age and HRL histologic results were used in the model, as were text features from the biopsy pathologic report. Results One thousand six HRLs were identified, with a cancer upgrade rate of 11.4% (115 of 1006). A machine learning random forest model was developed with 671 HRLs and tested with an independent set of 335 HRLs. Among the most important traditional features were age and HRL histologic results (eg, atypical ductal hyperplasia). An important text feature from the pathologic reports was "severely atypical." Instead of surgical excision of all HRLs, if those categorized with the model to be at low risk for upgrade were surveilled and the remainder were excised, then 97.4% (37 of 38) of malignancies would have been diagnosed at surgery, and 30.6% (91 of 297) of surgeries of benign lesions could have been avoided. Conclusion This study provides proof of concept that a machine learning model can be applied to predict the risk of upgrade of HRLs to cancer. Use of this model could decrease unnecessary surgery by nearly one-third and could help guide clinical decision making with regard to surveillance versus surgical excision of HRLs. © RSNA, 2017.

ADMET Evaluation in Drug Discovery. 18. Reliable Prediction of Chemical-Induced Urinary Tract Toxicity by Boosting Machine Learning Approaches.

PubMed

Lei, Tailong; Sun, Huiyong; Kang, Yu; Zhu, Feng; Liu, Hui; Zhou, Wenfang; Wang, Zhe; Li, Dan; Li, Youyong; Hou, Tingjun

2017-11-06

Xenobiotic chemicals and their metabolites are mainly excreted out of our bodies by the urinary tract through the urine. Chemical-induced urinary tract toxicity is one of the main reasons that cause failure during drug development, and it is a common adverse event for medications, natural supplements, and environmental chemicals. Despite its importance, there are only a few in silico models for assessing urinary tract toxicity for a large number of compounds with diverse chemical structures. Here, we developed a series of qualitative and quantitative structure-activity relationship (QSAR) models for predicting urinary tract toxicity. In our study, the recursive feature elimination method incorporated with random forests (RFE-RF) was used for dimension reduction, and then eight machine learning approaches were used for QSAR modeling, i.e., relevance vector machine (RVM), support vector machine (SVM), regularized random forest (RRF), C5.0 trees, eXtreme gradient boosting (XGBoost), AdaBoost.M1, SVM boosting (SVMBoost), and RVM boosting (RVMBoost). For building classification models, the synthetic minority oversampling technique was used to handle the imbalance data set problem. Among all the machine learning approaches, SVMBoost based on the RBF kernel achieves both the best quantitative (q ext 2 = 0.845) and qualitative predictions for the test set (MCC of 0.787, AUC of 0.893, sensitivity of 89.6%, specificity of 94.1%, and global accuracy of 90.8%). The application domains were then analyzed, and all of the tested chemicals fall within the application domain coverage. We also examined the structure features of the chemicals with large prediction errors. In brief, both the regression and classification models developed by the SVMBoost approach have reliable prediction capability for assessing chemical-induced urinary tract toxicity.

Improving the Spatial Prediction of Soil Organic Carbon Stocks in a Complex Tropical Mountain Landscape by Methodological Specifications in Machine Learning Approaches

PubMed Central

Schmidt, Johannes; Glaser, Bruno

2016-01-01

Tropical forests are significant carbon sinks and their soils’ carbon storage potential is immense. However, little is known about the soil organic carbon (SOC) stocks of tropical mountain areas whose complex soil-landscape and difficult accessibility pose a challenge to spatial analysis. The choice of methodology for spatial prediction is of high importance to improve the expected poor model results in case of low predictor-response correlations. Four aspects were considered to improve model performance in predicting SOC stocks of the organic layer of a tropical mountain forest landscape: Different spatial predictor settings, predictor selection strategies, various machine learning algorithms and model tuning. Five machine learning algorithms: random forests, artificial neural networks, multivariate adaptive regression splines, boosted regression trees and support vector machines were trained and tuned to predict SOC stocks from predictors derived from a digital elevation model and satellite image. Topographical predictors were calculated with a GIS search radius of 45 to 615 m. Finally, three predictor selection strategies were applied to the total set of 236 predictors. All machine learning algorithms—including the model tuning and predictor selection—were compared via five repetitions of a tenfold cross-validation. The boosted regression tree algorithm resulted in the overall best model. SOC stocks ranged between 0.2 to 17.7 kg m-2, displaying a huge variability with diffuse insolation and curvatures of different scale guiding the spatial pattern. Predictor selection and model tuning improved the models’ predictive performance in all five machine learning algorithms. The rather low number of selected predictors favours forward compared to backward selection procedures. Choosing predictors due to their indiviual performance was vanquished by the two procedures which accounted for predictor interaction. PMID:27128736

Improving the Spatial Prediction of Soil Organic Carbon Stocks in a Complex Tropical Mountain Landscape by Methodological Specifications in Machine Learning Approaches.

PubMed

Ließ, Mareike; Schmidt, Johannes; Glaser, Bruno

2016-01-01

Tropical forests are significant carbon sinks and their soils' carbon storage potential is immense. However, little is known about the soil organic carbon (SOC) stocks of tropical mountain areas whose complex soil-landscape and difficult accessibility pose a challenge to spatial analysis. The choice of methodology for spatial prediction is of high importance to improve the expected poor model results in case of low predictor-response correlations. Four aspects were considered to improve model performance in predicting SOC stocks of the organic layer of a tropical mountain forest landscape: Different spatial predictor settings, predictor selection strategies, various machine learning algorithms and model tuning. Five machine learning algorithms: random forests, artificial neural networks, multivariate adaptive regression splines, boosted regression trees and support vector machines were trained and tuned to predict SOC stocks from predictors derived from a digital elevation model and satellite image. Topographical predictors were calculated with a GIS search radius of 45 to 615 m. Finally, three predictor selection strategies were applied to the total set of 236 predictors. All machine learning algorithms-including the model tuning and predictor selection-were compared via five repetitions of a tenfold cross-validation. The boosted regression tree algorithm resulted in the overall best model. SOC stocks ranged between 0.2 to 17.7 kg m-2, displaying a huge variability with diffuse insolation and curvatures of different scale guiding the spatial pattern. Predictor selection and model tuning improved the models' predictive performance in all five machine learning algorithms. The rather low number of selected predictors favours forward compared to backward selection procedures. Choosing predictors due to their indiviual performance was vanquished by the two procedures which accounted for predictor interaction.

Underwater image enhancement through depth estimation based on random forest

NASA Astrophysics Data System (ADS)

Tai, Shen-Chuan; Tsai, Ting-Chou; Huang, Jyun-Han

2017-11-01

Light absorption and scattering in underwater environments can result in low-contrast images with a distinct color cast. This paper proposes a systematic framework for the enhancement of underwater images. Light transmission is estimated using the random forest algorithm. RGB values, luminance, color difference, blurriness, and the dark channel are treated as features in training and estimation. Transmission is calculated using an ensemble machine learning algorithm to deal with a variety of conditions encountered in underwater environments. A color compensation and contrast enhancement algorithm based on depth information was also developed with the aim of improving the visual quality of underwater images. Experimental results demonstrate that the proposed scheme outperforms existing methods with regard to subjective visual quality as well as objective measurements.

Support vector machine in machine condition monitoring and fault diagnosis

NASA Astrophysics Data System (ADS)

Widodo, Achmad; Yang, Bo-Suk

2007-08-01

Recently, the issue of machine condition monitoring and fault diagnosis as a part of maintenance system became global due to the potential advantages to be gained from reduced maintenance costs, improved productivity and increased machine availability. This paper presents a survey of machine condition monitoring and fault diagnosis using support vector machine (SVM). It attempts to summarize and review the recent research and developments of SVM in machine condition monitoring and diagnosis. Numerous methods have been developed based on intelligent systems such as artificial neural network, fuzzy expert system, condition-based reasoning, random forest, etc. However, the use of SVM for machine condition monitoring and fault diagnosis is still rare. SVM has excellent performance in generalization so it can produce high accuracy in classification for machine condition monitoring and diagnosis. Until 2006, the use of SVM in machine condition monitoring and fault diagnosis is tending to develop towards expertise orientation and problem-oriented domain. Finally, the ability to continually change and obtain a novel idea for machine condition monitoring and fault diagnosis using SVM will be future works.

Characterizing stand-level forest canopy cover and height using Landsat time series, samples of airborne LiDAR, and the Random Forest algorithm

NASA Astrophysics Data System (ADS)

Ahmed, Oumer S.; Franklin, Steven E.; Wulder, Michael A.; White, Joanne C.

2015-03-01

Many forest management activities, including the development of forest inventories, require spatially detailed forest canopy cover and height data. Among the various remote sensing technologies, LiDAR (Light Detection and Ranging) offers the most accurate and consistent means for obtaining reliable canopy structure measurements. A potential solution to reduce the cost of LiDAR data, is to integrate transects (samples) of LiDAR data with frequently acquired and spatially comprehensive optical remotely sensed data. Although multiple regression is commonly used for such modeling, often it does not fully capture the complex relationships between forest structure variables. This study investigates the potential of Random Forest (RF), a machine learning technique, to estimate LiDAR measured canopy structure using a time series of Landsat imagery. The study is implemented over a 2600 ha area of industrially managed coastal temperate forests on Vancouver Island, British Columbia, Canada. We implemented a trajectory-based approach to time series analysis that generates time since disturbance (TSD) and disturbance intensity information for each pixel and we used this information to stratify the forest land base into two strata: mature forests and young forests. Canopy cover and height for three forest classes (i.e. mature, young and mature and young (combined)) were modeled separately using multiple regression and Random Forest (RF) techniques. For all forest classes, the RF models provided improved estimates relative to the multiple regression models. The lowest validation error was obtained for the mature forest strata in a RF model (R2 = 0.88, RMSE = 2.39 m and bias = -0.16 for canopy height; R2 = 0.72, RMSE = 0.068% and bias = -0.0049 for canopy cover). This study demonstrates the value of using disturbance and successional history to inform estimates of canopy structure and obtain improved estimates of forest canopy cover and height using the RF algorithm.

Forecasting Solar Flares Using Magnetogram-based Predictors and Machine Learning

NASA Astrophysics Data System (ADS)

Florios, Kostas; Kontogiannis, Ioannis; Park, Sung-Hong; Guerra, Jordan A.; Benvenuto, Federico; Bloomfield, D. Shaun; Georgoulis, Manolis K.

2018-02-01

We propose a forecasting approach for solar flares based on data from Solar Cycle 24, taken by the Helioseismic and Magnetic Imager (HMI) on board the Solar Dynamics Observatory (SDO) mission. In particular, we use the Space-weather HMI Active Region Patches (SHARP) product that facilitates cut-out magnetograms of solar active regions (AR) in the Sun in near-realtime (NRT), taken over a five-year interval (2012 - 2016). Our approach utilizes a set of thirteen predictors, which are not included in the SHARP metadata, extracted from line-of-sight and vector photospheric magnetograms. We exploit several machine learning (ML) and conventional statistics techniques to predict flares of peak magnitude {>} M1 and {>} C1 within a 24 h forecast window. The ML methods used are multi-layer perceptrons (MLP), support vector machines (SVM), and random forests (RF). We conclude that random forests could be the prediction technique of choice for our sample, with the second-best method being multi-layer perceptrons, subject to an entropy objective function. A Monte Carlo simulation showed that the best-performing method gives accuracy ACC=0.93(0.00), true skill statistic TSS=0.74(0.02), and Heidke skill score HSS=0.49(0.01) for {>} M1 flare prediction with probability threshold 15% and ACC=0.84(0.00), TSS=0.60(0.01), and HSS=0.59(0.01) for {>} C1 flare prediction with probability threshold 35%.

Pre-operative prediction of surgical morbidity in children: comparison of five statistical models.

PubMed

Cooper, Jennifer N; Wei, Lai; Fernandez, Soledad A; Minneci, Peter C; Deans, Katherine J

2015-02-01

The accurate prediction of surgical risk is important to patients and physicians. Logistic regression (LR) models are typically used to estimate these risks. However, in the fields of data mining and machine-learning, many alternative classification and prediction algorithms have been developed. This study aimed to compare the performance of LR to several data mining algorithms for predicting 30-day surgical morbidity in children. We used the 2012 National Surgical Quality Improvement Program-Pediatric dataset to compare the performance of (1) a LR model that assumed linearity and additivity (simple LR model) (2) a LR model incorporating restricted cubic splines and interactions (flexible LR model) (3) a support vector machine, (4) a random forest and (5) boosted classification trees for predicting surgical morbidity. The ensemble-based methods showed significantly higher accuracy, sensitivity, specificity, PPV, and NPV than the simple LR model. However, none of the models performed better than the flexible LR model in terms of the aforementioned measures or in model calibration or discrimination. Support vector machines, random forests, and boosted classification trees do not show better performance than LR for predicting pediatric surgical morbidity. After further validation, the flexible LR model derived in this study could be used to assist with clinical decision-making based on patient-specific surgical risks. Copyright © 2014 Elsevier Ltd. All rights reserved.

Using GPS to evaluate productivity and performance of forest machine systems

Treesearch

Steven E. Taylor; Timothy P. McDonald; Matthew W. Veal; Ton E. Grift

2001-01-01

This paper reviews recent research and operational applications of using GPS as a tool to help monitor the locations, travel patterns, performance, and productivity of forest machines. The accuracy of dynamic GPS data collected on forest machines under different levels of forest canopy is reviewed first. Then, the paper focuses on the use of GPS for monitoring forest...

Data Analytics for SAR

DOE Office of Scientific and Technical Information (OSTI.GOV)

Murphy, David Patrick; Calef, Matthew Thomas

We assess the ability of variants of anomalous change detection (ACD) to identify human activity associated with large outdoor music festivals as they are seen from synthetic aperture radar (SAR) imagery collected by the Sentinel-1 satellite constellation. We found that, with appropriate feature vectors, ACD using random-forest machine learning was most effective at identifying changes associated with the human activity.

Virtual screening by a new Clustering-based Weighted Similarity Extreme Learning Machine approach

PubMed Central

Kudisthalert, Wasu

2018-01-01

Machine learning techniques are becoming popular in virtual screening tasks. One of the powerful machine learning algorithms is Extreme Learning Machine (ELM) which has been applied to many applications and has recently been applied to virtual screening. We propose the Weighted Similarity ELM (WS-ELM) which is based on a single layer feed-forward neural network in a conjunction of 16 different similarity coefficients as activation function in the hidden layer. It is known that the performance of conventional ELM is not robust due to random weight selection in the hidden layer. Thus, we propose a Clustering-based WS-ELM (CWS-ELM) that deterministically assigns weights by utilising clustering algorithms i.e. k-means clustering and support vector clustering. The experiments were conducted on one of the most challenging datasets–Maximum Unbiased Validation Dataset–which contains 17 activity classes carefully selected from PubChem. The proposed algorithms were then compared with other machine learning techniques such as support vector machine, random forest, and similarity searching. The results show that CWS-ELM in conjunction with support vector clustering yields the best performance when utilised together with Sokal/Sneath(1) coefficient. Furthermore, ECFP_6 fingerprint presents the best results in our framework compared to the other types of fingerprints, namely ECFP_4, FCFP_4, and FCFP_6. PMID:29652912

Computer-Aided Screening of Conjugated Polymers for Organic Solar Cell: Classification by Random Forest.

PubMed

Nagasawa, Shinji; Al-Naamani, Eman; Saeki, Akinori

2018-05-17

Owing to the diverse chemical structures, organic photovoltaic (OPV) applications with a bulk heterojunction framework have greatly evolved over the last two decades, which has produced numerous organic semiconductors exhibiting improved power conversion efficiencies (PCEs). Despite the recent fast progress in materials informatics and data science, data-driven molecular design of OPV materials remains challenging. We report a screening of conjugated molecules for polymer-fullerene OPV applications by supervised learning methods (artificial neural network (ANN) and random forest (RF)). Approximately 1000 experimental parameters including PCE, molecular weight, and electronic properties are manually collected from the literature and subjected to machine learning with digitized chemical structures. Contrary to the low correlation coefficient in ANN, RF yields an acceptable accuracy, which is twice that of random classification. We demonstrate the application of RF screening for the design, synthesis, and characterization of a conjugated polymer, which facilitates a rapid development of optoelectronic materials.

Hubble Tarantula Treasury Project - VI. Identification of Pre-Main-Sequence Stars using Machine Learning techniques

NASA Astrophysics Data System (ADS)

Ksoll, Victor F.; Gouliermis, Dimitrios A.; Klessen, Ralf S.; Grebel, Eva K.; Sabbi, Elena; Anderson, Jay; Lennon, Daniel J.; Cignoni, Michele; de Marchi, Guido; Smith, Linda J.; Tosi, Monica; van der Marel, Roeland P.

2018-05-01

The Hubble Tarantula Treasury Project (HTTP) has provided an unprecedented photometric coverage of the entire star-burst region of 30 Doradus down to the half Solar mass limit. We use the deep stellar catalogue of HTTP to identify all the pre-main-sequence (PMS) stars of the region, i.e., stars that have not started their lives on the main-sequence yet. The photometric distinction of these stars from the more evolved populations is not a trivial task due to several factors that alter their colour-magnitude diagram positions. The identification of PMS stars requires, thus, sophisticated statistical methods. We employ Machine Learning Classification techniques on the HTTP survey of more than 800,000 sources to identify the PMS stellar content of the observed field. Our methodology consists of 1) carefully selecting the most probable low-mass PMS stellar population of the star-forming cluster NGC2070, 2) using this sample to train classification algorithms to build a predictive model for PMS stars, and 3) applying this model in order to identify the most probable PMS content across the entire Tarantula Nebula. We employ Decision Tree, Random Forest and Support Vector Machine classifiers to categorise the stars as PMS and Non-PMS. The Random Forest and Support Vector Machine provided the most accurate models, predicting about 20,000 sources with a candidateship probability higher than 50 percent, and almost 10,000 PMS candidates with a probability higher than 95 percent. This is the richest and most accurate photometric catalogue of extragalactic PMS candidates across the extent of a whole star-forming complex.

Supervised machine learning for analysing spectra of exoplanetary atmospheres

NASA Astrophysics Data System (ADS)

Márquez-Neila, Pablo; Fisher, Chloe; Sznitman, Raphael; Heng, Kevin

2018-06-01

The use of machine learning is becoming ubiquitous in astronomy1-3, but remains rare in the study of the atmospheres of exoplanets. Given the spectrum of an exoplanetary atmosphere, a multi-parameter space is swept through in real time to find the best-fit model4-6. Known as atmospheric retrieval, this technique originates in the Earth and planetary sciences7. Such methods are very time-consuming, and by necessity there is a compromise between physical and chemical realism and computational feasibility. Machine learning has previously been used to determine which molecules to include in the model, but the retrieval itself was still performed using standard methods8. Here, we report an adaptation of the `random forest' method of supervised machine learning9,10, trained on a precomputed grid of atmospheric models, which retrieves full posterior distributions of the abundances of molecules and the cloud opacity. The use of a precomputed grid allows a large part of the computational burden to be shifted offline. We demonstrate our technique on a transmission spectrum of the hot gas-giant exoplanet WASP-12b using a five-parameter model (temperature, a constant cloud opacity and the volume mixing ratios or relative abundances of molecules of water, ammonia and hydrogen cyanide)11. We obtain results consistent with the standard nested-sampling retrieval method. We also estimate the sensitivity of the measured spectrum to the model parameters, and we are able to quantify the information content of the spectrum. Our method can be straightforwardly applied using more sophisticated atmospheric models to interpret an ensemble of spectra without having to retrain the random forest.

Machine-learning-based classification of real-time tissue elastography for hepatic fibrosis in patients with chronic hepatitis B.

PubMed

Chen, Yang; Luo, Yan; Huang, Wei; Hu, Die; Zheng, Rong-Qin; Cong, Shu-Zhen; Meng, Fan-Kun; Yang, Hong; Lin, Hong-Jun; Sun, Yan; Wang, Xiu-Yan; Wu, Tao; Ren, Jie; Pei, Shu-Fang; Zheng, Ying; He, Yun; Hu, Yu; Yang, Na; Yan, Hongmei

2017-10-01

Hepatic fibrosis is a common middle stage of the pathological processes of chronic liver diseases. Clinical intervention during the early stages of hepatic fibrosis can slow the development of liver cirrhosis and reduce the risk of developing liver cancer. Performing a liver biopsy, the gold standard for viral liver disease management, has drawbacks such as invasiveness and a relatively high sampling error rate. Real-time tissue elastography (RTE), one of the most recently developed technologies, might be promising imaging technology because it is both noninvasive and provides accurate assessments of hepatic fibrosis. However, determining the stage of liver fibrosis from RTE images in a clinic is a challenging task. In this study, in contrast to the previous liver fibrosis index (LFI) method, which predicts the stage of diagnosis using RTE images and multiple regression analysis, we employed four classical classifiers (i.e., Support Vector Machine, Naïve Bayes, Random Forest and K-Nearest Neighbor) to build a decision-support system to improve the hepatitis B stage diagnosis performance. Eleven RTE image features were obtained from 513 subjects who underwent liver biopsies in this multicenter collaborative research. The experimental results showed that the adopted classifiers significantly outperformed the LFI method and that the Random Forest(RF) classifier provided the highest average accuracy among the four machine algorithms. This result suggests that sophisticated machine-learning methods can be powerful tools for evaluating the stage of hepatic fibrosis and show promise for clinical applications. Copyright © 2017 Elsevier Ltd. All rights reserved.

Integrating human and machine intelligence in galaxy morphology classification tasks

NASA Astrophysics Data System (ADS)

Beck, Melanie R.; Scarlata, Claudia; Fortson, Lucy F.; Lintott, Chris J.; Simmons, B. D.; Galloway, Melanie A.; Willett, Kyle W.; Dickinson, Hugh; Masters, Karen L.; Marshall, Philip J.; Wright, Darryl

2018-06-01

Quantifying galaxy morphology is a challenging yet scientifically rewarding task. As the scale of data continues to increase with upcoming surveys, traditional classification methods will struggle to handle the load. We present a solution through an integration of visual and automated classifications, preserving the best features of both human and machine. We demonstrate the effectiveness of such a system through a re-analysis of visual galaxy morphology classifications collected during the Galaxy Zoo 2 (GZ2) project. We reprocess the top-level question of the GZ2 decision tree with a Bayesian classification aggregation algorithm dubbed SWAP, originally developed for the Space Warps gravitational lens project. Through a simple binary classification scheme, we increase the classification rate nearly 5-fold classifying 226 124 galaxies in 92 d of GZ2 project time while reproducing labels derived from GZ2 classification data with 95.7 per cent accuracy. We next combine this with a Random Forest machine learning algorithm that learns on a suite of non-parametric morphology indicators widely used for automated morphologies. We develop a decision engine that delegates tasks between human and machine and demonstrate that the combined system provides at least a factor of 8 increase in the classification rate, classifying 210 803 galaxies in just 32 d of GZ2 project time with 93.1 per cent accuracy. As the Random Forest algorithm requires a minimal amount of computational cost, this result has important implications for galaxy morphology identification tasks in the era of Euclid and other large-scale surveys.

Predicting diabetes mellitus using SMOTE and ensemble machine learning approach: The Henry Ford ExercIse Testing (FIT) project.

PubMed

Alghamdi, Manal; Al-Mallah, Mouaz; Keteyian, Steven; Brawner, Clinton; Ehrman, Jonathan; Sakr, Sherif

2017-01-01

Machine learning is becoming a popular and important approach in the field of medical research. In this study, we investigate the relative performance of various machine learning methods such as Decision Tree, Naïve Bayes, Logistic Regression, Logistic Model Tree and Random Forests for predicting incident diabetes using medical records of cardiorespiratory fitness. In addition, we apply different techniques to uncover potential predictors of diabetes. This FIT project study used data of 32,555 patients who are free of any known coronary artery disease or heart failure who underwent clinician-referred exercise treadmill stress testing at Henry Ford Health Systems between 1991 and 2009 and had a complete 5-year follow-up. At the completion of the fifth year, 5,099 of those patients have developed diabetes. The dataset contained 62 attributes classified into four categories: demographic characteristics, disease history, medication use history, and stress test vital signs. We developed an Ensembling-based predictive model using 13 attributes that were selected based on their clinical importance, Multiple Linear Regression, and Information Gain Ranking methods. The negative effect of the imbalance class of the constructed model was handled by Synthetic Minority Oversampling Technique (SMOTE). The overall performance of the predictive model classifier was improved by the Ensemble machine learning approach using the Vote method with three Decision Trees (Naïve Bayes Tree, Random Forest, and Logistic Model Tree) and achieved high accuracy of prediction (AUC = 0.92). The study shows the potential of ensembling and SMOTE approaches for predicting incident diabetes using cardiorespiratory fitness data.

Using machine learning for sequence-level automated MRI protocol selection in neuroradiology.

PubMed

Brown, Andrew D; Marotta, Thomas R

2018-05-01

Incorrect imaging protocol selection can lead to important clinical findings being missed, contributing to both wasted health care resources and patient harm. We present a machine learning method for analyzing the unstructured text of clinical indications and patient demographics from magnetic resonance imaging (MRI) orders to automatically protocol MRI procedures at the sequence level. We compared 3 machine learning models - support vector machine, gradient boosting machine, and random forest - to a baseline model that predicted the most common protocol for all observations in our test set. The gradient boosting machine model significantly outperformed the baseline and demonstrated the best performance of the 3 models in terms of accuracy (95%), precision (86%), recall (80%), and Hamming loss (0.0487). This demonstrates the feasibility of automating sequence selection by applying machine learning to MRI orders. Automated sequence selection has important safety, quality, and financial implications and may facilitate improvements in the quality and safety of medical imaging service delivery.

Forest Cover Estimation in Ireland Using Radar Remote Sensing: A Comparative Analysis of Forest Cover Assessment Methodologies.

PubMed

Devaney, John; Barrett, Brian; Barrett, Frank; Redmond, John; O Halloran, John

2015-01-01

Quantification of spatial and temporal changes in forest cover is an essential component of forest monitoring programs. Due to its cloud free capability, Synthetic Aperture Radar (SAR) is an ideal source of information on forest dynamics in countries with near-constant cloud-cover. However, few studies have investigated the use of SAR for forest cover estimation in landscapes with highly sparse and fragmented forest cover. In this study, the potential use of L-band SAR for forest cover estimation in two regions (Longford and Sligo) in Ireland is investigated and compared to forest cover estimates derived from three national (Forestry2010, Prime2, National Forest Inventory), one pan-European (Forest Map 2006) and one global forest cover (Global Forest Change) product. Two machine-learning approaches (Random Forests and Extremely Randomised Trees) are evaluated. Both Random Forests and Extremely Randomised Trees classification accuracies were high (98.1-98.5%), with differences between the two classifiers being minimal (<0.5%). Increasing levels of post classification filtering led to a decrease in estimated forest area and an increase in overall accuracy of SAR-derived forest cover maps. All forest cover products were evaluated using an independent validation dataset. For the Longford region, the highest overall accuracy was recorded with the Forestry2010 dataset (97.42%) whereas in Sligo, highest overall accuracy was obtained for the Prime2 dataset (97.43%), although accuracies of SAR-derived forest maps were comparable. Our findings indicate that spaceborne radar could aid inventories in regions with low levels of forest cover in fragmented landscapes. The reduced accuracies observed for the global and pan-continental forest cover maps in comparison to national and SAR-derived forest maps indicate that caution should be exercised when applying these datasets for national reporting.

Forest Cover Estimation in Ireland Using Radar Remote Sensing: A Comparative Analysis of Forest Cover Assessment Methodologies

PubMed Central

Devaney, John; Barrett, Brian; Barrett, Frank; Redmond, John; O`Halloran, John

2015-01-01

Quantification of spatial and temporal changes in forest cover is an essential component of forest monitoring programs. Due to its cloud free capability, Synthetic Aperture Radar (SAR) is an ideal source of information on forest dynamics in countries with near-constant cloud-cover. However, few studies have investigated the use of SAR for forest cover estimation in landscapes with highly sparse and fragmented forest cover. In this study, the potential use of L-band SAR for forest cover estimation in two regions (Longford and Sligo) in Ireland is investigated and compared to forest cover estimates derived from three national (Forestry2010, Prime2, National Forest Inventory), one pan-European (Forest Map 2006) and one global forest cover (Global Forest Change) product. Two machine-learning approaches (Random Forests and Extremely Randomised Trees) are evaluated. Both Random Forests and Extremely Randomised Trees classification accuracies were high (98.1–98.5%), with differences between the two classifiers being minimal (<0.5%). Increasing levels of post classification filtering led to a decrease in estimated forest area and an increase in overall accuracy of SAR-derived forest cover maps. All forest cover products were evaluated using an independent validation dataset. For the Longford region, the highest overall accuracy was recorded with the Forestry2010 dataset (97.42%) whereas in Sligo, highest overall accuracy was obtained for the Prime2 dataset (97.43%), although accuracies of SAR-derived forest maps were comparable. Our findings indicate that spaceborne radar could aid inventories in regions with low levels of forest cover in fragmented landscapes. The reduced accuracies observed for the global and pan-continental forest cover maps in comparison to national and SAR-derived forest maps indicate that caution should be exercised when applying these datasets for national reporting. PMID:26262681

Machine rates for selected forest harvesting machines

Treesearch

R.W. Brinker; J. Kinard; Robert Rummer; B. Lanford

2002-01-01

Very little new literature has been published on the subject of machine rates and machine cost analysis since 1989 when the Alabama Agricultural Experiment Station Circular 296, Machine Rates for Selected Forest Harvesting Machines, was originally published. Many machines discussed in the original publication have undergone substantial changes in various aspects, not...

Comparison of four machine learning methods for object-oriented change detection in high-resolution satellite imagery

NASA Astrophysics Data System (ADS)

Bai, Ting; Sun, Kaimin; Deng, Shiquan; Chen, Yan

2018-03-01

High resolution image change detection is one of the key technologies of remote sensing application, which is of great significance for resource survey, environmental monitoring, fine agriculture, military mapping and battlefield environment detection. In this paper, for high-resolution satellite imagery, Random Forest (RF), Support Vector Machine (SVM), Deep belief network (DBN), and Adaboost models were established to verify the possibility of different machine learning applications in change detection. In order to compare detection accuracy of four machine learning Method, we applied these four machine learning methods for two high-resolution images. The results shows that SVM has higher overall accuracy at small samples compared to RF, Adaboost, and DBN for binary and from-to change detection. With the increase in the number of samples, RF has higher overall accuracy compared to Adaboost, SVM and DBN.

Predicting the dissolution kinetics of silicate glasses using machine learning

NASA Astrophysics Data System (ADS)

Anoop Krishnan, N. M.; Mangalathu, Sujith; Smedskjaer, Morten M.; Tandia, Adama; Burton, Henry; Bauchy, Mathieu

2018-05-01

Predicting the dissolution rates of silicate glasses in aqueous conditions is a complex task as the underlying mechanism(s) remain poorly understood and the dissolution kinetics can depend on a large number of intrinsic and extrinsic factors. Here, we assess the potential of data-driven models based on machine learning to predict the dissolution rates of various aluminosilicate glasses exposed to a wide range of solution pH values, from acidic to caustic conditions. Four classes of machine learning methods are investigated, namely, linear regression, support vector machine regression, random forest, and artificial neural network. We observe that, although linear methods all fail to describe the dissolution kinetics, the artificial neural network approach offers excellent predictions, thanks to its inherent ability to handle non-linear data. Overall, we suggest that a more extensive use of machine learning approaches could significantly accelerate the design of novel glasses with tailored properties.

A random forest algorithm for nowcasting of intense precipitation events

NASA Astrophysics Data System (ADS)

Das, Saurabh; Chakraborty, Rohit; Maitra, Animesh

2017-09-01

Automatic nowcasting of convective initiation and thunderstorms has potential applications in several sectors including aviation planning and disaster management. In this paper, random forest based machine learning algorithm is tested for nowcasting of convective rain with a ground based radiometer. Brightness temperatures measured at 14 frequencies (7 frequencies in 22-31 GHz band and 7 frequencies in 51-58 GHz bands) are utilized as the inputs of the model. The lower frequency band is associated to the water vapor absorption whereas the upper frequency band relates to the oxygen absorption and hence, provide information on the temperature and humidity of the atmosphere. Synthetic minority over-sampling technique is used to balance the data set and 10-fold cross validation is used to assess the performance of the model. Results indicate that random forest algorithm with fixed alarm generation time of 30 min and 60 min performs quite well (probability of detection of all types of weather condition ∼90%) with low false alarms. It is, however, also observed that reducing the alarm generation time improves the threat score significantly and also decreases false alarms. The proposed model is found to be very sensitive to the boundary layer instability as indicated by the variable importance measure. The study shows the suitability of a random forest algorithm for nowcasting application utilizing a large number of input parameters from diverse sources and can be utilized in other forecasting problems.

Machine learning algorithms for the creation of clinical healthcare enterprise systems

NASA Astrophysics Data System (ADS)

Mandal, Indrajit

2017-10-01

Clinical recommender systems are increasingly becoming popular for improving modern healthcare systems. Enterprise systems are persuasively used for creating effective nurse care plans to provide nurse training, clinical recommendations and clinical quality control. A novel design of a reliable clinical recommender system based on multiple classifier system (MCS) is implemented. A hybrid machine learning (ML) ensemble based on random subspace method and random forest is presented. The performance accuracy and robustness of proposed enterprise architecture are quantitatively estimated to be above 99% and 97%, respectively (above 95% confidence interval). The study then extends to experimental analysis of the clinical recommender system with respect to the noisy data environment. The ranking of items in nurse care plan is demonstrated using machine learning algorithms (MLAs) to overcome the drawback of the traditional association rule method. The promising experimental results are compared against the sate-of-the-art approaches to highlight the advancement in recommendation technology. The proposed recommender system is experimentally validated using five benchmark clinical data to reinforce the research findings.

Using Classification and Regression Trees (CART) and random forests to analyze attrition: Results from two simulations.

PubMed

Hayes, Timothy; Usami, Satoshi; Jacobucci, Ross; McArdle, John J

2015-12-01

In this article, we describe a recent development in the analysis of attrition: using classification and regression trees (CART) and random forest methods to generate inverse sampling weights. These flexible machine learning techniques have the potential to capture complex nonlinear, interactive selection models, yet to our knowledge, their performance in the missing data analysis context has never been evaluated. To assess the potential benefits of these methods, we compare their performance with commonly employed multiple imputation and complete case techniques in 2 simulations. These initial results suggest that weights computed from pruned CART analyses performed well in terms of both bias and efficiency when compared with other methods. We discuss the implications of these findings for applied researchers. (c) 2015 APA, all rights reserved).

Using Classification and Regression Trees (CART) and Random Forests to Analyze Attrition: Results From Two Simulations

PubMed Central

Hayes, Timothy; Usami, Satoshi; Jacobucci, Ross; McArdle, John J.

2016-01-01

In this article, we describe a recent development in the analysis of attrition: using classification and regression trees (CART) and random forest methods to generate inverse sampling weights. These flexible machine learning techniques have the potential to capture complex nonlinear, interactive selection models, yet to our knowledge, their performance in the missing data analysis context has never been evaluated. To assess the potential benefits of these methods, we compare their performance with commonly employed multiple imputation and complete case techniques in 2 simulations. These initial results suggest that weights computed from pruned CART analyses performed well in terms of both bias and efficiency when compared with other methods. We discuss the implications of these findings for applied researchers. PMID:26389526

Differential privacy-based evaporative cooling feature selection and classification with relief-F and random forests.

PubMed

Le, Trang T; Simmons, W Kyle; Misaki, Masaya; Bodurka, Jerzy; White, Bill C; Savitz, Jonathan; McKinney, Brett A

2017-09-15

Classification of individuals into disease or clinical categories from high-dimensional biological data with low prediction error is an important challenge of statistical learning in bioinformatics. Feature selection can improve classification accuracy but must be incorporated carefully into cross-validation to avoid overfitting. Recently, feature selection methods based on differential privacy, such as differentially private random forests and reusable holdout sets, have been proposed. However, for domains such as bioinformatics, where the number of features is much larger than the number of observations p≫n , these differential privacy methods are susceptible to overfitting. We introduce private Evaporative Cooling, a stochastic privacy-preserving machine learning algorithm that uses Relief-F for feature selection and random forest for privacy preserving classification that also prevents overfitting. We relate the privacy-preserving threshold mechanism to a thermodynamic Maxwell-Boltzmann distribution, where the temperature represents the privacy threshold. We use the thermal statistical physics concept of Evaporative Cooling of atomic gases to perform backward stepwise privacy-preserving feature selection. On simulated data with main effects and statistical interactions, we compare accuracies on holdout and validation sets for three privacy-preserving methods: the reusable holdout, reusable holdout with random forest, and private Evaporative Cooling, which uses Relief-F feature selection and random forest classification. In simulations where interactions exist between attributes, private Evaporative Cooling provides higher classification accuracy without overfitting based on an independent validation set. In simulations without interactions, thresholdout with random forest and private Evaporative Cooling give comparable accuracies. We also apply these privacy methods to human brain resting-state fMRI data from a study of major depressive disorder. Code available at http://insilico.utulsa.edu/software/privateEC . brett-mckinney@utulsa.edu. Supplementary data are available at Bioinformatics online. © The Author (2017). Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com

Random forest regression modelling for forest aboveground biomass estimation using RISAT-1 PolSAR and terrestrial LiDAR data

NASA Astrophysics Data System (ADS)

Mangla, Rohit; Kumar, Shashi; Nandy, Subrata

2016-05-01

SAR and LiDAR remote sensing have already shown the potential of active sensors for forest parameter retrieval. SAR sensor in its fully polarimetric mode has an advantage to retrieve scattering property of different component of forest structure and LiDAR has the capability to measure structural information with very high accuracy. This study was focused on retrieval of forest aboveground biomass (AGB) using Terrestrial Laser Scanner (TLS) based point clouds and scattering property of forest vegetation obtained from decomposition modelling of RISAT-1 fully polarimetric SAR data. TLS data was acquired for 14 plots of Timli forest range, Uttarakhand, India. The forest area is dominated by Sal trees and random sampling with plot size of 0.1 ha (31.62m*31.62m) was adopted for TLS and field data collection. RISAT-1 data was processed to retrieve SAR data based variables and TLS point clouds based 3D imaging was done to retrieve LiDAR based variables. Surface scattering, double-bounce scattering, volume scattering, helix and wire scattering were the SAR based variables retrieved from polarimetric decomposition. Tree heights and stem diameters were used as LiDAR based variables retrieved from single tree vertical height and least square circle fit methods respectively. All the variables obtained for forest plots were used as an input in a machine learning based Random Forest Regression Model, which was developed in this study for forest AGB estimation. Modelled output for forest AGB showed reliable accuracy (RMSE = 27.68 t/ha) and a good coefficient of determination (0.63) was obtained through the linear regression between modelled AGB and field-estimated AGB. The sensitivity analysis showed that the model was more sensitive for the major contributed variables (stem diameter and volume scattering) and these variables were measured from two different remote sensing techniques. This study strongly recommends the integration of SAR and LiDAR data for forest AGB estimation.

Prediction of Short-Distance Aerial Movement of Phakopsora pachyrhizi Urediniospores Using Machine Learning.

PubMed

Wen, L; Bowen, C R; Hartman, G L

2017-10-01

Dispersal of urediniospores by wind is the primary means of spread for Phakopsora pachyrhizi, the cause of soybean rust. Our research focused on the short-distance movement of urediniospores from within the soybean canopy and up to 61 m from field-grown rust-infected soybean plants. Environmental variables were used to develop and compare models including the least absolute shrinkage and selection operator regression, zero-inflated Poisson/regular Poisson regression, random forest, and neural network to describe deposition of urediniospores collected in passive and active traps. All four models identified distance of trap from source, humidity, temperature, wind direction, and wind speed as the five most important variables influencing short-distance movement of urediniospores. The random forest model provided the best predictions, explaining 76.1 and 86.8% of the total variation in the passive- and active-trap datasets, respectively. The prediction accuracy based on the correlation coefficient (r) between predicted values and the true values were 0.83 (P < 0.0001) and 0.94 (P < 0.0001) for the passive and active trap datasets, respectively. Overall, multiple machine learning techniques identified the most important variables to make the most accurate predictions of movement of P. pachyrhizi urediniospores short-distance.

Remote sensing-based measurement of Living Environment Deprivation: Improving classical approaches with machine learning

PubMed Central

2017-01-01

This paper provides evidence on the usefulness of very high spatial resolution (VHR) imagery in gathering socioeconomic information in urban settlements. We use land cover, spectral, structure and texture features extracted from a Google Earth image of Liverpool (UK) to evaluate their potential to predict Living Environment Deprivation at a small statistical area level. We also contribute to the methodological literature on the estimation of socioeconomic indices with remote-sensing data by introducing elements from modern machine learning. In addition to classical approaches such as Ordinary Least Squares (OLS) regression and a spatial lag model, we explore the potential of the Gradient Boost Regressor and Random Forests to improve predictive performance and accuracy. In addition to novel predicting methods, we also introduce tools for model interpretation and evaluation such as feature importance and partial dependence plots, or cross-validation. Our results show that Random Forest proved to be the best model with an R2 of around 0.54, followed by Gradient Boost Regressor with 0.5. Both the spatial lag model and the OLS fall behind with significantly lower performances of 0.43 and 0.3, respectively. PMID:28464010

Remote sensing-based measurement of Living Environment Deprivation: Improving classical approaches with machine learning.

PubMed

Arribas-Bel, Daniel; Patino, Jorge E; Duque, Juan C

2017-01-01

This paper provides evidence on the usefulness of very high spatial resolution (VHR) imagery in gathering socioeconomic information in urban settlements. We use land cover, spectral, structure and texture features extracted from a Google Earth image of Liverpool (UK) to evaluate their potential to predict Living Environment Deprivation at a small statistical area level. We also contribute to the methodological literature on the estimation of socioeconomic indices with remote-sensing data by introducing elements from modern machine learning. In addition to classical approaches such as Ordinary Least Squares (OLS) regression and a spatial lag model, we explore the potential of the Gradient Boost Regressor and Random Forests to improve predictive performance and accuracy. In addition to novel predicting methods, we also introduce tools for model interpretation and evaluation such as feature importance and partial dependence plots, or cross-validation. Our results show that Random Forest proved to be the best model with an R2 of around 0.54, followed by Gradient Boost Regressor with 0.5. Both the spatial lag model and the OLS fall behind with significantly lower performances of 0.43 and 0.3, respectively.

Preliminary investigation of human exhaled breath for tuberculosis diagnosis by multidimensional gas chromatography - Time of flight mass spectrometry and machine learning.

PubMed

Beccaria, Marco; Mellors, Theodore R; Petion, Jacky S; Rees, Christiaan A; Nasir, Mavra; Systrom, Hannah K; Sairistil, Jean W; Jean-Juste, Marc-Antoine; Rivera, Vanessa; Lavoile, Kerline; Severe, Patrice; Pape, Jean W; Wright, Peter F; Hill, Jane E

2018-02-01

Tuberculosis (TB) remains a global public health malady that claims almost 1.8 million lives annually. Diagnosis of TB represents perhaps one of the most challenging aspects of tuberculosis control. Gold standards for diagnosis of active TB (culture and nucleic acid amplification) are sputum-dependent, however, in up to a third of TB cases, an adequate biological sputum sample is not readily available. The analysis of exhaled breath, as an alternative to sputum-dependent tests, has the potential to provide a simple, fast, and non-invasive, and ready-available diagnostic service that could positively change TB detection. Human breath has been evaluated in the setting of active tuberculosis using thermal desorption-comprehensive two-dimensional gas chromatography-time of flight mass spectrometry methodology. From the entire spectrum of volatile metabolites in breath, three random forest machine learning models were applied leading to the generation of a panel of 46 breath features. The twenty-two common features within each random forest model used were selected as a set that could distinguish subjects with confirmed pulmonary M. tuberculosis infection and people with other pathologies than TB. Copyright © 2018 Elsevier B.V. All rights reserved.

Automatic detection of atrial fibrillation in cardiac vibration signals.

PubMed

Brueser, C; Diesel, J; Zink, M D H; Winter, S; Schauerte, P; Leonhardt, S

2013-01-01

We present a study on the feasibility of the automatic detection of atrial fibrillation (AF) from cardiac vibration signals (ballistocardiograms/BCGs) recorded by unobtrusive bedmounted sensors. The proposed system is intended as a screening and monitoring tool in home-healthcare applications and not as a replacement for ECG-based methods used in clinical environments. Based on BCG data recorded in a study with 10 AF patients, we evaluate and rank seven popular machine learning algorithms (naive Bayes, linear and quadratic discriminant analysis, support vector machines, random forests as well as bagged and boosted trees) for their performance in separating 30 s long BCG epochs into one of three classes: sinus rhythm, atrial fibrillation, and artifact. For each algorithm, feature subsets of a set of statistical time-frequency-domain and time-domain features were selected based on the mutual information between features and class labels as well as first- and second-order interactions among features. The classifiers were evaluated on a set of 856 epochs by means of 10-fold cross-validation. The best algorithm (random forests) achieved a Matthews correlation coefficient, mean sensitivity, and mean specificity of 0.921, 0.938, and 0.982, respectively.

Beyond the hype: deep neural networks outperform established methods using a ChEMBL bioactivity benchmark set.

PubMed

Lenselink, Eelke B; Ten Dijke, Niels; Bongers, Brandon; Papadatos, George; van Vlijmen, Herman W T; Kowalczyk, Wojtek; IJzerman, Adriaan P; van Westen, Gerard J P

2017-08-14

The increase of publicly available bioactivity data in recent years has fueled and catalyzed research in chemogenomics, data mining, and modeling approaches. As a direct result, over the past few years a multitude of different methods have been reported and evaluated, such as target fishing, nearest neighbor similarity-based methods, and Quantitative Structure Activity Relationship (QSAR)-based protocols. However, such studies are typically conducted on different datasets, using different validation strategies, and different metrics. In this study, different methods were compared using one single standardized dataset obtained from ChEMBL, which is made available to the public, using standardized metrics (BEDROC and Matthews Correlation Coefficient). Specifically, the performance of Naïve Bayes, Random Forests, Support Vector Machines, Logistic Regression, and Deep Neural Networks was assessed using QSAR and proteochemometric (PCM) methods. All methods were validated using both a random split validation and a temporal validation, with the latter being a more realistic benchmark of expected prospective execution. Deep Neural Networks are the top performing classifiers, highlighting the added value of Deep Neural Networks over other more conventional methods. Moreover, the best method ('DNN_PCM') performed significantly better at almost one standard deviation higher than the mean performance. Furthermore, Multi-task and PCM implementations were shown to improve performance over single task Deep Neural Networks. Conversely, target prediction performed almost two standard deviations under the mean performance. Random Forests, Support Vector Machines, and Logistic Regression performed around mean performance. Finally, using an ensemble of DNNs, alongside additional tuning, enhanced the relative performance by another 27% (compared with unoptimized 'DNN_PCM'). Here, a standardized set to test and evaluate different machine learning algorithms in the context of multi-task learning is offered by providing the data and the protocols. Graphical Abstract .

Retrieval of aerosol optical depth from surface solar radiation measurements using machine learning algorithms, non-linear regression and a radiative transfer-based look-up table

NASA Astrophysics Data System (ADS)

Huttunen, Jani; Kokkola, Harri; Mielonen, Tero; Esa Juhani Mononen, Mika; Lipponen, Antti; Reunanen, Juha; Vilhelm Lindfors, Anders; Mikkonen, Santtu; Erkki Juhani Lehtinen, Kari; Kouremeti, Natalia; Bais, Alkiviadis; Niska, Harri; Arola, Antti

2016-07-01

In order to have a good estimate of the current forcing by anthropogenic aerosols, knowledge on past aerosol levels is needed. Aerosol optical depth (AOD) is a good measure for aerosol loading. However, dedicated measurements of AOD are only available from the 1990s onward. One option to lengthen the AOD time series beyond the 1990s is to retrieve AOD from surface solar radiation (SSR) measurements taken with pyranometers. In this work, we have evaluated several inversion methods designed for this task. We compared a look-up table method based on radiative transfer modelling, a non-linear regression method and four machine learning methods (Gaussian process, neural network, random forest and support vector machine) with AOD observations carried out with a sun photometer at an Aerosol Robotic Network (AERONET) site in Thessaloniki, Greece. Our results show that most of the machine learning methods produce AOD estimates comparable to the look-up table and non-linear regression methods. All of the applied methods produced AOD values that corresponded well to the AERONET observations with the lowest correlation coefficient value being 0.87 for the random forest method. While many of the methods tended to slightly overestimate low AODs and underestimate high AODs, neural network and support vector machine showed overall better correspondence for the whole AOD range. The differences in producing both ends of the AOD range seem to be caused by differences in the aerosol composition. High AODs were in most cases those with high water vapour content which might affect the aerosol single scattering albedo (SSA) through uptake of water into aerosols. Our study indicates that machine learning methods benefit from the fact that they do not constrain the aerosol SSA in the retrieval, whereas the LUT method assumes a constant value for it. This would also mean that machine learning methods could have potential in reproducing AOD from SSR even though SSA would have changed during the observation period.

Stable Atlas-based Mapped Prior (STAMP) machine-learning segmentation for multicenter large-scale MRI data.

PubMed

Kim, Eun Young; Magnotta, Vincent A; Liu, Dawei; Johnson, Hans J

2014-09-01

Machine learning (ML)-based segmentation methods are a common technique in the medical image processing field. In spite of numerous research groups that have investigated ML-based segmentation frameworks, there remains unanswered aspects of performance variability for the choice of two key components: ML algorithm and intensity normalization. This investigation reveals that the choice of those elements plays a major part in determining segmentation accuracy and generalizability. The approach we have used in this study aims to evaluate relative benefits of the two elements within a subcortical MRI segmentation framework. Experiments were conducted to contrast eight machine-learning algorithm configurations and 11 normalization strategies for our brain MR segmentation framework. For the intensity normalization, a Stable Atlas-based Mapped Prior (STAMP) was utilized to take better account of contrast along boundaries of structures. Comparing eight machine learning algorithms on down-sampled segmentation MR data, it was obvious that a significant improvement was obtained using ensemble-based ML algorithms (i.e., random forest) or ANN algorithms. Further investigation between these two algorithms also revealed that the random forest results provided exceptionally good agreement with manual delineations by experts. Additional experiments showed that the effect of STAMP-based intensity normalization also improved the robustness of segmentation for multicenter data sets. The constructed framework obtained good multicenter reliability and was successfully applied on a large multicenter MR data set (n>3000). Less than 10% of automated segmentations were recommended for minimal expert intervention. These results demonstrate the feasibility of using the ML-based segmentation tools for processing large amount of multicenter MR images. We demonstrated dramatically different result profiles in segmentation accuracy according to the choice of ML algorithm and intensity normalization chosen. Copyright © 2014 Elsevier Inc. All rights reserved.

Automated diagnosis of myositis from muscle ultrasound: Exploring the use of machine learning and deep learning methods

PubMed Central

Burlina, Philippe; Billings, Seth; Joshi, Neil

2017-01-01

Objective To evaluate the use of ultrasound coupled with machine learning (ML) and deep learning (DL) techniques for automated or semi-automated classification of myositis. Methods Eighty subjects comprised of 19 with inclusion body myositis (IBM), 14 with polymyositis (PM), 14 with dermatomyositis (DM), and 33 normal (N) subjects were included in this study, where 3214 muscle ultrasound images of 7 muscles (observed bilaterally) were acquired. We considered three problems of classification including (A) normal vs. affected (DM, PM, IBM); (B) normal vs. IBM patients; and (C) IBM vs. other types of myositis (DM or PM). We studied the use of an automated DL method using deep convolutional neural networks (DL-DCNNs) for diagnostic classification and compared it with a semi-automated conventional ML method based on random forests (ML-RF) and “engineered” features. We used the known clinical diagnosis as the gold standard for evaluating performance of muscle classification. Results The performance of the DL-DCNN method resulted in accuracies ± standard deviation of 76.2% ± 3.1% for problem (A), 86.6% ± 2.4% for (B) and 74.8% ± 3.9% for (C), while the ML-RF method led to accuracies of 72.3% ± 3.3% for problem (A), 84.3% ± 2.3% for (B) and 68.9% ± 2.5% for (C). Conclusions This study demonstrates the application of machine learning methods for automatically or semi-automatically classifying inflammatory muscle disease using muscle ultrasound. Compared to the conventional random forest machine learning method used here, which has the drawback of requiring manual delineation of muscle/fat boundaries, DCNN-based classification by and large improved the accuracies in all classification problems while providing a fully automated approach to classification. PMID:28854220

Automated diagnosis of myositis from muscle ultrasound: Exploring the use of machine learning and deep learning methods.

PubMed

Burlina, Philippe; Billings, Seth; Joshi, Neil; Albayda, Jemima

2017-01-01

To evaluate the use of ultrasound coupled with machine learning (ML) and deep learning (DL) techniques for automated or semi-automated classification of myositis. Eighty subjects comprised of 19 with inclusion body myositis (IBM), 14 with polymyositis (PM), 14 with dermatomyositis (DM), and 33 normal (N) subjects were included in this study, where 3214 muscle ultrasound images of 7 muscles (observed bilaterally) were acquired. We considered three problems of classification including (A) normal vs. affected (DM, PM, IBM); (B) normal vs. IBM patients; and (C) IBM vs. other types of myositis (DM or PM). We studied the use of an automated DL method using deep convolutional neural networks (DL-DCNNs) for diagnostic classification and compared it with a semi-automated conventional ML method based on random forests (ML-RF) and "engineered" features. We used the known clinical diagnosis as the gold standard for evaluating performance of muscle classification. The performance of the DL-DCNN method resulted in accuracies ± standard deviation of 76.2% ± 3.1% for problem (A), 86.6% ± 2.4% for (B) and 74.8% ± 3.9% for (C), while the ML-RF method led to accuracies of 72.3% ± 3.3% for problem (A), 84.3% ± 2.3% for (B) and 68.9% ± 2.5% for (C). This study demonstrates the application of machine learning methods for automatically or semi-automatically classifying inflammatory muscle disease using muscle ultrasound. Compared to the conventional random forest machine learning method used here, which has the drawback of requiring manual delineation of muscle/fat boundaries, DCNN-based classification by and large improved the accuracies in all classification problems while providing a fully automated approach to classification.

Classification of suicide attempters in schizophrenia using sociocultural and clinical features: A machine learning approach.

PubMed

Hettige, Nuwan C; Nguyen, Thai Binh; Yuan, Chen; Rajakulendran, Thanara; Baddour, Jermeen; Bhagwat, Nikhil; Bani-Fatemi, Ali; Voineskos, Aristotle N; Mallar Chakravarty, M; De Luca, Vincenzo

2017-07-01

Suicide is a major concern for those afflicted by schizophrenia. Identifying patients at the highest risk for future suicide attempts remains a complex problem for psychiatric interventions. Machine learning models allow for the integration of many risk factors in order to build an algorithm that predicts which patients are likely to attempt suicide. Currently it is unclear how to integrate previously identified risk factors into a clinically relevant predictive tool to estimate the probability of a patient with schizophrenia for attempting suicide. We conducted a cross-sectional assessment on a sample of 345 participants diagnosed with schizophrenia spectrum disorders. Suicide attempters and non-attempters were clearly identified using the Columbia Suicide Severity Rating Scale (C-SSRS) and the Beck Suicide Ideation Scale (BSS). We developed four classification algorithms using a regularized regression, random forest, elastic net and support vector machine models with sociocultural and clinical variables as features to train the models. All classification models performed similarly in identifying suicide attempters and non-attempters. Our regularized logistic regression model demonstrated an accuracy of 67% and an area under the curve (AUC) of 0.71, while the random forest model demonstrated 66% accuracy and an AUC of 0.67. Support vector classifier (SVC) model demonstrated an accuracy of 67% and an AUC of 0.70, and the elastic net model demonstrated and accuracy of 65% and an AUC of 0.71. Machine learning algorithms offer a relatively successful method for incorporating many clinical features to predict individuals at risk for future suicide attempts. Increased performance of these models using clinically relevant variables offers the potential to facilitate early treatment and intervention to prevent future suicide attempts. Copyright © 2017 Elsevier Inc. All rights reserved.

Machine Learning Algorithm Predicts Cardiac Resynchronization Therapy Outcomes: Lessons From the COMPANION Trial.

PubMed

Kalscheur, Matthew M; Kipp, Ryan T; Tattersall, Matthew C; Mei, Chaoqun; Buhr, Kevin A; DeMets, David L; Field, Michael E; Eckhardt, Lee L; Page, C David

2018-01-01

Cardiac resynchronization therapy (CRT) reduces morbidity and mortality in heart failure patients with reduced left ventricular function and intraventricular conduction delay. However, individual outcomes vary significantly. This study sought to use a machine learning algorithm to develop a model to predict outcomes after CRT. Models were developed with machine learning algorithms to predict all-cause mortality or heart failure hospitalization at 12 months post-CRT in the COMPANION trial (Comparison of Medical Therapy, Pacing, and Defibrillation in Heart Failure). The best performing model was developed with the random forest algorithm. The ability of this model to predict all-cause mortality or heart failure hospitalization and all-cause mortality alone was compared with discrimination obtained using a combination of bundle branch block morphology and QRS duration. In the 595 patients with CRT-defibrillator in the COMPANION trial, 105 deaths occurred (median follow-up, 15.7 months). The survival difference across subgroups differentiated by bundle branch block morphology and QRS duration did not reach significance ( P =0.08). The random forest model produced quartiles of patients with an 8-fold difference in survival between those with the highest and lowest predicted probability for events (hazard ratio, 7.96; P <0.0001). The model also discriminated the risk of the composite end point of all-cause mortality or heart failure hospitalization better than subgroups based on bundle branch block morphology and QRS duration. In the COMPANION trial, a machine learning algorithm produced a model that predicted clinical outcomes after CRT. Applied before device implant, this model may better differentiate outcomes over current clinical discriminators and improve shared decision-making with patients. © 2018 American Heart Association, Inc.

Data-driven mapping of the potential mountain permafrost distribution.

PubMed

Deluigi, Nicola; Lambiel, Christophe; Kanevski, Mikhail

2017-07-15

Existing mountain permafrost distribution models generally offer a good overview of the potential extent of this phenomenon at a regional scale. They are however not always able to reproduce the high spatial discontinuity of permafrost at the micro-scale (scale of a specific landform; ten to several hundreds of meters). To overcome this lack, we tested an alternative modelling approach using three classification algorithms belonging to statistics and machine learning: Logistic regression, Support Vector Machines and Random forests. These supervised learning techniques infer a classification function from labelled training data (pixels of permafrost absence and presence) with the aim of predicting the permafrost occurrence where it is unknown. The research was carried out in a 588km 2 area of the Western Swiss Alps. Permafrost evidences were mapped from ortho-image interpretation (rock glacier inventorying) and field data (mainly geoelectrical and thermal data). The relationship between selected permafrost evidences and permafrost controlling factors was computed with the mentioned techniques. Classification performances, assessed with AUROC, range between 0.81 for Logistic regression, 0.85 with Support Vector Machines and 0.88 with Random forests. The adopted machine learning algorithms have demonstrated to be efficient for permafrost distribution modelling thanks to consistent results compared to the field reality. The high resolution of the input dataset (10m) allows elaborating maps at the micro-scale with a modelled permafrost spatial distribution less optimistic than classic spatial models. Moreover, the probability output of adopted algorithms offers a more precise overview of the potential distribution of mountain permafrost than proposing simple indexes of the permafrost favorability. These encouraging results also open the way to new possibilities of permafrost data analysis and mapping. Copyright © 2017 Elsevier B.V. All rights reserved.

Enhancing interpretability of automatically extracted machine learning features: application to a RBM-Random Forest system on brain lesion segmentation.

PubMed

Pereira, Sérgio; Meier, Raphael; McKinley, Richard; Wiest, Roland; Alves, Victor; Silva, Carlos A; Reyes, Mauricio

2018-02-01

Machine learning systems are achieving better performances at the cost of becoming increasingly complex. However, because of that, they become less interpretable, which may cause some distrust by the end-user of the system. This is especially important as these systems are pervasively being introduced to critical domains, such as the medical field. Representation Learning techniques are general methods for automatic feature computation. Nevertheless, these techniques are regarded as uninterpretable "black boxes". In this paper, we propose a methodology to enhance the interpretability of automatically extracted machine learning features. The proposed system is composed of a Restricted Boltzmann Machine for unsupervised feature learning, and a Random Forest classifier, which are combined to jointly consider existing correlations between imaging data, features, and target variables. We define two levels of interpretation: global and local. The former is devoted to understanding if the system learned the relevant relations in the data correctly, while the later is focused on predictions performed on a voxel- and patient-level. In addition, we propose a novel feature importance strategy that considers both imaging data and target variables, and we demonstrate the ability of the approach to leverage the interpretability of the obtained representation for the task at hand. We evaluated the proposed methodology in brain tumor segmentation and penumbra estimation in ischemic stroke lesions. We show the ability of the proposed methodology to unveil information regarding relationships between imaging modalities and extracted features and their usefulness for the task at hand. In both clinical scenarios, we demonstrate that the proposed methodology enhances the interpretability of automatically learned features, highlighting specific learning patterns that resemble how an expert extracts relevant data from medical images. Copyright © 2017 Elsevier B.V. All rights reserved.

Machine-z: Rapid Machine-Learned Redshift Indicator for Swift Gamma-Ray Bursts

NASA Technical Reports Server (NTRS)

Ukwatta, T. N.; Wozniak, P. R.; Gehrels, N.

2016-01-01

Studies of high-redshift gamma-ray bursts (GRBs) provide important information about the early Universe such as the rates of stellar collapsars and mergers, the metallicity content, constraints on the re-ionization period, and probes of the Hubble expansion. Rapid selection of high-z candidates from GRB samples reported in real time by dedicated space missions such as Swift is the key to identifying the most distant bursts before the optical afterglow becomes too dim to warrant a good spectrum. Here, we introduce 'machine-z', a redshift prediction algorithm and a 'high-z' classifier for Swift GRBs based on machine learning. Our method relies exclusively on canonical data commonly available within the first few hours after the GRB trigger. Using a sample of 284 bursts with measured redshifts, we trained a randomized ensemble of decision trees (random forest) to perform both regression and classification. Cross-validated performance studies show that the correlation coefficient between machine-z predictions and the true redshift is nearly 0.6. At the same time, our high-z classifier can achieve 80 per cent recall of true high-redshift bursts, while incurring a false positive rate of 20 per cent. With 40 per cent false positive rate the classifier can achieve approximately 100 per cent recall. The most reliable selection of high-redshift GRBs is obtained by combining predictions from both the high-z classifier and the machine-z regressor.

Application of Machine-Learning Models to Predict Tacrolimus Stable Dose in Renal Transplant Recipients

NASA Astrophysics Data System (ADS)

Tang, Jie; Liu, Rong; Zhang, Yue-Li; Liu, Mou-Ze; Hu, Yong-Fang; Shao, Ming-Jie; Zhu, Li-Jun; Xin, Hua-Wen; Feng, Gui-Wen; Shang, Wen-Jun; Meng, Xiang-Guang; Zhang, Li-Rong; Ming, Ying-Zi; Zhang, Wei

2017-02-01

Tacrolimus has a narrow therapeutic window and considerable variability in clinical use. Our goal was to compare the performance of multiple linear regression (MLR) and eight machine learning techniques in pharmacogenetic algorithm-based prediction of tacrolimus stable dose (TSD) in a large Chinese cohort. A total of 1,045 renal transplant patients were recruited, 80% of which were randomly selected as the “derivation cohort” to develop dose-prediction algorithm, while the remaining 20% constituted the “validation cohort” to test the final selected algorithm. MLR, artificial neural network (ANN), regression tree (RT), multivariate adaptive regression splines (MARS), boosted regression tree (BRT), support vector regression (SVR), random forest regression (RFR), lasso regression (LAR) and Bayesian additive regression trees (BART) were applied and their performances were compared in this work. Among all the machine learning models, RT performed best in both derivation [0.71 (0.67-0.76)] and validation cohorts [0.73 (0.63-0.82)]. In addition, the ideal rate of RT was 4% higher than that of MLR. To our knowledge, this is the first study to use machine learning models to predict TSD, which will further facilitate personalized medicine in tacrolimus administration in the future.

Global Positioning System (GPS) and Geographic Information System (GIS) analysis of mobile harvesting equipment and sediment delivery to streams during forest harvest operations on steep terrain: Experimental design

Treesearch

Daniel Bowker; Jeff Stringer; Chris Barton; Songlin Fei

2011-01-01

Sediment mobilized by forest harvest machine traffic contributes substantially to the degradation of headwater stream systems. This study monitored forest harvest machine traffic to analyze how it affects sediment delivery to stream channels. Harvest machines were outfitted with global positioning system (GPS) dataloggers, recording machine movements and working status...

Introducing two Random Forest based methods for cloud detection in remote sensing images

NASA Astrophysics Data System (ADS)

Ghasemian, Nafiseh; Akhoondzadeh, Mehdi

2018-07-01

Cloud detection is a necessary phase in satellite images processing to retrieve the atmospheric and lithospheric parameters. Currently, some cloud detection methods based on Random Forest (RF) model have been proposed but they do not consider both spectral and textural characteristics of the image. Furthermore, they have not been tested in the presence of snow/ice. In this paper, we introduce two RF based algorithms, Feature Level Fusion Random Forest (FLFRF) and Decision Level Fusion Random Forest (DLFRF) to incorporate visible, infrared (IR) and thermal spectral and textural features (FLFRF) including Gray Level Co-occurrence Matrix (GLCM) and Robust Extended Local Binary Pattern (RELBP_CI) or visible, IR and thermal classifiers (DLFRF) for highly accurate cloud detection on remote sensing images. FLFRF first fuses visible, IR and thermal features. Thereafter, it uses the RF model to classify pixels to cloud, snow/ice and background or thick cloud, thin cloud and background. DLFRF considers visible, IR and thermal features (both spectral and textural) separately and inserts each set of features to RF model. Then, it holds vote matrix of each run of the model. Finally, it fuses the classifiers using the majority vote method. To demonstrate the effectiveness of the proposed algorithms, 10 Terra MODIS and 15 Landsat 8 OLI/TIRS images with different spatial resolutions are used in this paper. Quantitative analyses are based on manually selected ground truth data. Results show that after adding RELBP_CI to input feature set cloud detection accuracy improves. Also, the average cloud kappa values of FLFRF and DLFRF on MODIS images (1 and 0.99) are higher than other machine learning methods, Linear Discriminate Analysis (LDA), Classification And Regression Tree (CART), K Nearest Neighbor (KNN) and Support Vector Machine (SVM) (0.96). The average snow/ice kappa values of FLFRF and DLFRF on MODIS images (1 and 0.85) are higher than other traditional methods. The quantitative values on Landsat 8 images show similar trend. Consequently, while SVM and K-nearest neighbor show overestimation in predicting cloud and snow/ice pixels, our Random Forest (RF) based models can achieve higher cloud, snow/ice kappa values on MODIS and thin cloud, thick cloud and snow/ice kappa values on Landsat 8 images. Our algorithms predict both thin and thick cloud on Landsat 8 images while the existing cloud detection algorithm, Fmask cannot discriminate them. Compared to the state-of-the-art methods, our algorithms have acquired higher average cloud and snow/ice kappa values for different spatial resolutions.

Analysis of landslide hazard area in Ludian earthquake based on Random Forests

NASA Astrophysics Data System (ADS)

Xie, J.-C.; Liu, R.; Li, H.-W.; Lai, Z.-L.

2015-04-01

With the development of machine learning theory, more and more algorithms are evaluated for seismic landslides. After the Ludian earthquake, the research team combine with the special geological structure in Ludian area and the seismic filed exploration results, selecting SLOPE(PODU); River distance(HL); Fault distance(DC); Seismic Intensity(LD) and Digital Elevation Model(DEM), the normalized difference vegetation index(NDVI) which based on remote sensing images as evaluation factors. But the relationships among these factors are fuzzy, there also exists heavy noise and high-dimensional, we introduce the random forest algorithm to tolerate these difficulties and get the evaluation result of Ludian landslide areas, in order to verify the accuracy of the result, using the ROC graphs for the result evaluation standard, AUC covers an area of 0.918, meanwhile, the random forest's generalization error rate decreases with the increase of the classification tree to the ideal 0.08 by using Out Of Bag(OOB) Estimation. Studying the final landslides inversion results, paper comes to a statistical conclusion that near 80% of the whole landslides and dilapidations are in areas with high susceptibility and moderate susceptibility, showing the forecast results are reasonable and adopted.

Predicting Coastal Flood Severity using Random Forest Algorithm

NASA Astrophysics Data System (ADS)

Sadler, J. M.; Goodall, J. L.; Morsy, M. M.; Spencer, K.

2017-12-01

Coastal floods have become more common recently and are predicted to further increase in frequency and severity due to sea level rise. Predicting floods in coastal cities can be difficult due to the number of environmental and geographic factors which can influence flooding events. Built stormwater infrastructure and irregular urban landscapes add further complexity. This paper demonstrates the use of machine learning algorithms in predicting street flood occurrence in an urban coastal setting. The model is trained and evaluated using data from Norfolk, Virginia USA from September 2010 - October 2016. Rainfall, tide levels, water table levels, and wind conditions are used as input variables. Street flooding reports made by city workers after named and unnamed storm events, ranging from 1-159 reports per event, are the model output. Results show that Random Forest provides predictive power in estimating the number of flood occurrences given a set of environmental conditions with an out-of-bag root mean squared error of 4.3 flood reports and a mean absolute error of 0.82 flood reports. The Random Forest algorithm performed much better than Poisson regression. From the Random Forest model, total daily rainfall was by far the most important factor in flood occurrence prediction, followed by daily low tide and daily higher high tide. The model demonstrated here could be used to predict flood severity based on forecast rainfall and tide conditions and could be further enhanced using more complete street flooding data for model training.

Machine learning methods in chemoinformatics

PubMed Central

Mitchell, John B O

2014-01-01

Machine learning algorithms are generally developed in computer science or adjacent disciplines and find their way into chemical modeling by a process of diffusion. Though particular machine learning methods are popular in chemoinformatics and quantitative structure–activity relationships (QSAR), many others exist in the technical literature. This discussion is methods-based and focused on some algorithms that chemoinformatics researchers frequently use. It makes no claim to be exhaustive. We concentrate on methods for supervised learning, predicting the unknown property values of a test set of instances, usually molecules, based on the known values for a training set. Particularly relevant approaches include Artificial Neural Networks, Random Forest, Support Vector Machine, k-Nearest Neighbors and naïve Bayes classifiers. WIREs Comput Mol Sci 2014, 4:468–481. How to cite this article: WIREs Comput Mol Sci 2014, 4:468–481. doi:10.1002/wcms.1183 PMID:25285160

A Comparison of Machine Learning Approaches for Corn Yield Estimation

NASA Astrophysics Data System (ADS)

Kim, N.; Lee, Y. W.

2017-12-01

Machine learning is an efficient empirical method for classification and prediction, and it is another approach to crop yield estimation. The objective of this study is to estimate corn yield in the Midwestern United States by employing the machine learning approaches such as the support vector machine (SVM), random forest (RF), and deep neural networks (DNN), and to perform the comprehensive comparison for their results. We constructed the database using satellite images from MODIS, the climate data of PRISM climate group, and GLDAS soil moisture data. In addition, to examine the seasonal sensitivities of corn yields, two period groups were set up: May to September (MJJAS) and July and August (JA). In overall, the DNN showed the highest accuracies in term of the correlation coefficient for the two period groups. The differences between our predictions and USDA yield statistics were about 10-11 %.

Non-invasive estimate of blood glucose and blood pressure from a photoplethysmograph by means of machine learning techniques.

PubMed

Monte-Moreno, Enric

2011-10-01

This work presents a system for a simultaneous non-invasive estimate of the blood glucose level (BGL) and the systolic (SBP) and diastolic (DBP) blood pressure, using a photoplethysmograph (PPG) and machine learning techniques. The method is independent of the person whose values are being measured and does not need calibration over time or subjects. The architecture of the system consists of a photoplethysmograph sensor, an activity detection module, a signal processing module that extracts features from the PPG waveform, and a machine learning algorithm that estimates the SBP, DBP and BGL values. The idea that underlies the system is that there is functional relationship between the shape of the PPG waveform and the blood pressure and glucose levels. As described in this paper we tested this method on 410 individuals without performing any personalized calibration. The results were computed after cross validation. The machine learning techniques tested were: ridge linear regression, a multilayer perceptron neural network, support vector machines and random forests. The best results were obtained with the random forest technique. In the case of blood pressure, the resulting coefficients of determination for reference vs. prediction were R(SBP)(2)=0.91, R(DBP)(2)=0.89, and R(BGL)(2)=0.90. For the glucose estimation, distribution of the points on a Clarke error grid placed 87.7% of points in zone A, 10.3% in zone B, and 1.9% in zone D. Blood pressure values complied with the grade B protocol of the British Hypertension society. An effective system for estimate of blood glucose and blood pressure from a photoplethysmograph is presented. The main advantage of the system is that for clinical use it complies with the grade B protocol of the British Hypertension society for the blood pressure and only in 1.9% of the cases did not detect hypoglycemia or hyperglycemia. Copyright © 2011 Elsevier B.V. All rights reserved.

Application of Machine Learning Approaches for Protein-protein Interactions Prediction.

PubMed

Zhang, Mengying; Su, Qiang; Lu, Yi; Zhao, Manman; Niu, Bing

2017-01-01

Proteomics endeavors to study the structures, functions and interactions of proteins. Information of the protein-protein interactions (PPIs) helps to improve our knowledge of the functions and the 3D structures of proteins. Thus determining the PPIs is essential for the study of the proteomics. In this review, in order to study the application of machine learning in predicting PPI, some machine learning approaches such as support vector machine (SVM), artificial neural networks (ANNs) and random forest (RF) were selected, and the examples of its applications in PPIs were listed. SVM and RF are two commonly used methods. Nowadays, more researchers predict PPIs by combining more than two methods. This review presents the application of machine learning approaches in predicting PPI. Many examples of success in identification and prediction in the area of PPI prediction have been discussed, and the PPIs research is still in progress. Copyright© Bentham Science Publishers; For any queries, please email at epub@benthamscience.org.

Machine Learning Techniques for Prediction of Early Childhood Obesity.

PubMed

Dugan, T M; Mukhopadhyay, S; Carroll, A; Downs, S

2015-01-01

This paper aims to predict childhood obesity after age two, using only data collected prior to the second birthday by a clinical decision support system called CHICA. Analyses of six different machine learning methods: RandomTree, RandomForest, J48, ID3, Naïve Bayes, and Bayes trained on CHICA data show that an accurate, sensitive model can be created. Of the methods analyzed, the ID3 model trained on the CHICA dataset proved the best overall performance with accuracy of 85% and sensitivity of 89%. Additionally, the ID3 model had a positive predictive value of 84% and a negative predictive value of 88%. The structure of the tree also gives insight into the strongest predictors of future obesity in children. Many of the strongest predictors seen in the ID3 modeling of the CHICA dataset have been independently validated in the literature as correlated with obesity, thereby supporting the validity of the model. This study demonstrated that data from a production clinical decision support system can be used to build an accurate machine learning model to predict obesity in children after age two.

Prediction of the effect of formulation on the toxicity of chemicals.

PubMed

Mistry, Pritesh; Neagu, Daniel; Sanchez-Ruiz, Antonio; Trundle, Paul R; Vessey, Jonathan D; Gosling, John Paul

2017-01-01

Two approaches for the prediction of which of two vehicles will result in lower toxicity for anticancer agents are presented. Machine-learning models are developed using decision tree, random forest and partial least squares methodologies and statistical evidence is presented to demonstrate that they represent valid models. Separately, a clustering method is presented that allows the ordering of vehicles by the toxicity they show for chemically-related compounds.

Ensemble Methods for Classification of Physical Activities from Wrist Accelerometry.

PubMed

Chowdhury, Alok Kumar; Tjondronegoro, Dian; Chandran, Vinod; Trost, Stewart G

2017-09-01

To investigate whether the use of ensemble learning algorithms improve physical activity recognition accuracy compared to the single classifier algorithms, and to compare the classification accuracy achieved by three conventional ensemble machine learning methods (bagging, boosting, random forest) and a custom ensemble model comprising four algorithms commonly used for activity recognition (binary decision tree, k nearest neighbor, support vector machine, and neural network). The study used three independent data sets that included wrist-worn accelerometer data. For each data set, a four-step classification framework consisting of data preprocessing, feature extraction, normalization and feature selection, and classifier training and testing was implemented. For the custom ensemble, decisions from the single classifiers were aggregated using three decision fusion methods: weighted majority vote, naïve Bayes combination, and behavior knowledge space combination. Classifiers were cross-validated using leave-one subject out cross-validation and compared on the basis of average F1 scores. In all three data sets, ensemble learning methods consistently outperformed the individual classifiers. Among the conventional ensemble methods, random forest models provided consistently high activity recognition; however, the custom ensemble model using weighted majority voting demonstrated the highest classification accuracy in two of the three data sets. Combining multiple individual classifiers using conventional or custom ensemble learning methods can improve activity recognition accuracy from wrist-worn accelerometer data.

Random Forests for Global and Regional Crop Yield Predictions.

PubMed

Jeong, Jig Han; Resop, Jonathan P; Mueller, Nathaniel D; Fleisher, David H; Yun, Kyungdahm; Butler, Ethan E; Timlin, Dennis J; Shim, Kyo-Moon; Gerber, James S; Reddy, Vangimalla R; Kim, Soo-Hyung

2016-01-01

Accurate predictions of crop yield are critical for developing effective agricultural and food policies at the regional and global scales. We evaluated a machine-learning method, Random Forests (RF), for its ability to predict crop yield responses to climate and biophysical variables at global and regional scales in wheat, maize, and potato in comparison with multiple linear regressions (MLR) serving as a benchmark. We used crop yield data from various sources and regions for model training and testing: 1) gridded global wheat grain yield, 2) maize grain yield from US counties over thirty years, and 3) potato tuber and maize silage yield from the northeastern seaboard region. RF was found highly capable of predicting crop yields and outperformed MLR benchmarks in all performance statistics that were compared. For example, the root mean square errors (RMSE) ranged between 6 and 14% of the average observed yield with RF models in all test cases whereas these values ranged from 14% to 49% for MLR models. Our results show that RF is an effective and versatile machine-learning method for crop yield predictions at regional and global scales for its high accuracy and precision, ease of use, and utility in data analysis. RF may result in a loss of accuracy when predicting the extreme ends or responses beyond the boundaries of the training data.

Evaluating the statistical performance of less applied algorithms in classification of worldview-3 imagery data in an urbanized landscape

NASA Astrophysics Data System (ADS)

Ranaie, Mehrdad; Soffianian, Alireza; Pourmanafi, Saeid; Mirghaffari, Noorollah; Tarkesh, Mostafa

2018-03-01

In recent decade, analyzing the remotely sensed imagery is considered as one of the most common and widely used procedures in the environmental studies. In this case, supervised image classification techniques play a central role. Hence, taking a high resolution Worldview-3 over a mixed urbanized landscape in Iran, three less applied image classification methods including Bagged CART, Stochastic gradient boosting model and Neural network with feature extraction were tested and compared with two prevalent methods: random forest and support vector machine with linear kernel. To do so, each method was run ten time and three validation techniques was used to estimate the accuracy statistics consist of cross validation, independent validation and validation with total of train data. Moreover, using ANOVA and Tukey test, statistical difference significance between the classification methods was significantly surveyed. In general, the results showed that random forest with marginal difference compared to Bagged CART and stochastic gradient boosting model is the best performing method whilst based on independent validation there was no significant difference between the performances of classification methods. It should be finally noted that neural network with feature extraction and linear support vector machine had better processing speed than other.

DHSpred: support-vector-machine-based human DNase I hypersensitive sites prediction using the optimal features selected by random forest.

PubMed

Manavalan, Balachandran; Shin, Tae Hwan; Lee, Gwang

2018-01-05

DNase I hypersensitive sites (DHSs) are genomic regions that provide important information regarding the presence of transcriptional regulatory elements and the state of chromatin. Therefore, identifying DHSs in uncharacterized DNA sequences is crucial for understanding their biological functions and mechanisms. Although many experimental methods have been proposed to identify DHSs, they have proven to be expensive for genome-wide application. Therefore, it is necessary to develop computational methods for DHS prediction. In this study, we proposed a support vector machine (SVM)-based method for predicting DHSs, called DHSpred (DNase I Hypersensitive Site predictor in human DNA sequences), which was trained with 174 optimal features. The optimal combination of features was identified from a large set that included nucleotide composition and di- and trinucleotide physicochemical properties, using a random forest algorithm. DHSpred achieved a Matthews correlation coefficient and accuracy of 0.660 and 0.871, respectively, which were 3% higher than those of control SVM predictors trained with non-optimized features, indicating the efficiency of the feature selection method. Furthermore, the performance of DHSpred was superior to that of state-of-the-art predictors. An online prediction server has been developed to assist the scientific community, and is freely available at: http://www.thegleelab.org/DHSpred.html.

DHSpred: support-vector-machine-based human DNase I hypersensitive sites prediction using the optimal features selected by random forest

PubMed Central

Manavalan, Balachandran; Shin, Tae Hwan; Lee, Gwang

2018-01-01

DNase I hypersensitive sites (DHSs) are genomic regions that provide important information regarding the presence of transcriptional regulatory elements and the state of chromatin. Therefore, identifying DHSs in uncharacterized DNA sequences is crucial for understanding their biological functions and mechanisms. Although many experimental methods have been proposed to identify DHSs, they have proven to be expensive for genome-wide application. Therefore, it is necessary to develop computational methods for DHS prediction. In this study, we proposed a support vector machine (SVM)-based method for predicting DHSs, called DHSpred (DNase I Hypersensitive Site predictor in human DNA sequences), which was trained with 174 optimal features. The optimal combination of features was identified from a large set that included nucleotide composition and di- and trinucleotide physicochemical properties, using a random forest algorithm. DHSpred achieved a Matthews correlation coefficient and accuracy of 0.660 and 0.871, respectively, which were 3% higher than those of control SVM predictors trained with non-optimized features, indicating the efficiency of the feature selection method. Furthermore, the performance of DHSpred was superior to that of state-of-the-art predictors. An online prediction server has been developed to assist the scientific community, and is freely available at: http://www.thegleelab.org/DHSpred.html PMID:29416743

Tissue segmentation of computed tomography images using a Random Forest algorithm: a feasibility study

NASA Astrophysics Data System (ADS)

Polan, Daniel F.; Brady, Samuel L.; Kaufman, Robert A.

2016-09-01

There is a need for robust, fully automated whole body organ segmentation for diagnostic CT. This study investigates and optimizes a Random Forest algorithm for automated organ segmentation; explores the limitations of a Random Forest algorithm applied to the CT environment; and demonstrates segmentation accuracy in a feasibility study of pediatric and adult patients. To the best of our knowledge, this is the first study to investigate a trainable Weka segmentation (TWS) implementation using Random Forest machine-learning as a means to develop a fully automated tissue segmentation tool developed specifically for pediatric and adult examinations in a diagnostic CT environment. Current innovation in computed tomography (CT) is focused on radiomics, patient-specific radiation dose calculation, and image quality improvement using iterative reconstruction, all of which require specific knowledge of tissue and organ systems within a CT image. The purpose of this study was to develop a fully automated Random Forest classifier algorithm for segmentation of neck-chest-abdomen-pelvis CT examinations based on pediatric and adult CT protocols. Seven materials were classified: background, lung/internal air or gas, fat, muscle, solid organ parenchyma, blood/contrast enhanced fluid, and bone tissue using Matlab and the TWS plugin of FIJI. The following classifier feature filters of TWS were investigated: minimum, maximum, mean, and variance evaluated over a voxel radius of 2 n , (n from 0 to 4), along with noise reduction and edge preserving filters: Gaussian, bilateral, Kuwahara, and anisotropic diffusion. The Random Forest algorithm used 200 trees with 2 features randomly selected per node. The optimized auto-segmentation algorithm resulted in 16 image features including features derived from maximum, mean, variance Gaussian and Kuwahara filters. Dice similarity coefficient (DSC) calculations between manually segmented and Random Forest algorithm segmented images from 21 patient image sections, were analyzed. The automated algorithm produced segmentation of seven material classes with a median DSC of 0.86  ±  0.03 for pediatric patient protocols, and 0.85  ±  0.04 for adult patient protocols. Additionally, 100 randomly selected patient examinations were segmented and analyzed, and a mean sensitivity of 0.91 (range: 0.82-0.98), specificity of 0.89 (range: 0.70-0.98), and accuracy of 0.90 (range: 0.76-0.98) were demonstrated. In this study, we demonstrate that this fully automated segmentation tool was able to produce fast and accurate segmentation of the neck and trunk of the body over a wide range of patient habitus and scan parameters.

On the classification techniques in data mining for microarray data classification

NASA Astrophysics Data System (ADS)

Aydadenta, Husna; Adiwijaya

2018-03-01

Cancer is one of the deadly diseases, according to data from WHO by 2015 there are 8.8 million more deaths caused by cancer, and this will increase every year if not resolved earlier. Microarray data has become one of the most popular cancer-identification studies in the field of health, since microarray data can be used to look at levels of gene expression in certain cell samples that serve to analyze thousands of genes simultaneously. By using data mining technique, we can classify the sample of microarray data thus it can be identified with cancer or not. In this paper we will discuss some research using some data mining techniques using microarray data, such as Support Vector Machine (SVM), Artificial Neural Network (ANN), Naive Bayes, k-Nearest Neighbor (kNN), and C4.5, and simulation of Random Forest algorithm with technique of reduction dimension using Relief. The result of this paper show performance measure (accuracy) from classification algorithm (SVM, ANN, Naive Bayes, kNN, C4.5, and Random Forets).The results in this paper show the accuracy of Random Forest algorithm higher than other classification algorithms (Support Vector Machine (SVM), Artificial Neural Network (ANN), Naive Bayes, k-Nearest Neighbor (kNN), and C4.5). It is hoped that this paper can provide some information about the speed, accuracy, performance and computational cost generated from each Data Mining Classification Technique based on microarray data.

Learning About Climate and Atmospheric Models Through Machine Learning

NASA Astrophysics Data System (ADS)

Lucas, D. D.

2017-12-01

From the analysis of ensemble variability to improving simulation performance, machine learning algorithms can play a powerful role in understanding the behavior of atmospheric and climate models. To learn about model behavior, we create training and testing data sets through ensemble techniques that sample different model configurations and values of input parameters, and then use supervised machine learning to map the relationships between the inputs and outputs. Following this procedure, we have used support vector machines, random forests, gradient boosting and other methods to investigate a variety of atmospheric and climate model phenomena. We have used machine learning to predict simulation crashes, estimate the probability density function of climate sensitivity, optimize simulations of the Madden Julian oscillation, assess the impacts of weather and emissions uncertainty on atmospheric dispersion, and quantify the effects of model resolution changes on precipitation. This presentation highlights recent examples of our applications of machine learning to improve the understanding of climate and atmospheric models. This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344.

Machine- z: Rapid machine-learned redshift indicator for Swift gamma-ray bursts

DOE PAGES

Ukwatta, T. N.; Wozniak, P. R.; Gehrels, N.

2016-03-08

Studies of high-redshift gamma-ray bursts (GRBs) provide important information about the early Universe such as the rates of stellar collapsars and mergers, the metallicity content, constraints on the re-ionization period, and probes of the Hubble expansion. Rapid selection of high-z candidates from GRB samples reported in real time by dedicated space missions such as Swift is the key to identifying the most distant bursts before the optical afterglow becomes too dim to warrant a good spectrum. Here, we introduce ‘machine-z’, a redshift prediction algorithm and a ‘high-z’ classifier for Swift GRBs based on machine learning. Our method relies exclusively onmore » canonical data commonly available within the first few hours after the GRB trigger. Using a sample of 284 bursts with measured redshifts, we trained a randomized ensemble of decision trees (random forest) to perform both regression and classification. Cross-validated performance studies show that the correlation coefficient between machine-z predictions and the true redshift is nearly 0.6. At the same time, our high-z classifier can achieve 80 per cent recall of true high-redshift bursts, while incurring a false positive rate of 20 per cent. With 40 per cent false positive rate the classifier can achieve ~100 per cent recall. As a result, the most reliable selection of high-redshift GRBs is obtained by combining predictions from both the high-z classifier and the machine-z regressor.« less

Problematic internet use (PIU): Associations with the impulsive-compulsive spectrum. An application of machine learning in psychiatry.

PubMed

Ioannidis, Konstantinos; Chamberlain, Samuel R; Treder, Matthias S; Kiraly, Franz; Leppink, Eric W; Redden, Sarah A; Stein, Dan J; Lochner, Christine; Grant, Jon E

2016-12-01

Problematic internet use is common, functionally impairing, and in need of further study. Its relationship with obsessive-compulsive and impulsive disorders is unclear. Our objective was to evaluate whether problematic internet use can be predicted from recognised forms of impulsive and compulsive traits and symptomatology. We recruited volunteers aged 18 and older using media advertisements at two sites (Chicago USA, and Stellenbosch, South Africa) to complete an extensive online survey. State-of-the-art out-of-sample evaluation of machine learning predictive models was used, which included Logistic Regression, Random Forests and Naïve Bayes. Problematic internet use was identified using the Internet Addiction Test (IAT). 2006 complete cases were analysed, of whom 181 (9.0%) had moderate/severe problematic internet use. Using Logistic Regression and Naïve Bayes we produced a classification prediction with a receiver operating characteristic area under the curve (ROC-AUC) of 0.83 (SD 0.03) whereas using a Random Forests algorithm the prediction ROC-AUC was 0.84 (SD 0.03) [all three models superior to baseline models p < 0.0001]. The models showed robust transfer between the study sites in all validation sets [p < 0.0001]. Prediction of problematic internet use was possible using specific measures of impulsivity and compulsivity in a population of volunteers. Moreover, this study offers proof-of-concept in support of using machine learning in psychiatry to demonstrate replicability of results across geographically and culturally distinct settings. Copyright © 2016 The Author(s). Published by Elsevier Ltd.. All rights reserved.

Feature combination networks for the interpretation of statistical machine learning models: application to Ames mutagenicity.

PubMed

Webb, Samuel J; Hanser, Thierry; Howlin, Brendan; Krause, Paul; Vessey, Jonathan D

2014-03-25

A new algorithm has been developed to enable the interpretation of black box models. The developed algorithm is agnostic to learning algorithm and open to all structural based descriptors such as fragments, keys and hashed fingerprints. The algorithm has provided meaningful interpretation of Ames mutagenicity predictions from both random forest and support vector machine models built on a variety of structural fingerprints.A fragmentation algorithm is utilised to investigate the model's behaviour on specific substructures present in the query. An output is formulated summarising causes of activation and deactivation. The algorithm is able to identify multiple causes of activation or deactivation in addition to identifying localised deactivations where the prediction for the query is active overall. No loss in performance is seen as there is no change in the prediction; the interpretation is produced directly on the model's behaviour for the specific query. Models have been built using multiple learning algorithms including support vector machine and random forest. The models were built on public Ames mutagenicity data and a variety of fingerprint descriptors were used. These models produced a good performance in both internal and external validation with accuracies around 82%. The models were used to evaluate the interpretation algorithm. Interpretation was revealed that links closely with understood mechanisms for Ames mutagenicity. This methodology allows for a greater utilisation of the predictions made by black box models and can expedite further study based on the output for a (quantitative) structure activity model. Additionally the algorithm could be utilised for chemical dataset investigation and knowledge extraction/human SAR development.

How random is the random forest? Random forest algorithm on the service of structural imaging biomarkers for Alzheimer's disease: from Alzheimer's disease neuroimaging initiative (ADNI) database.

PubMed

Dimitriadis, Stavros I; Liparas, Dimitris

2018-06-01

Neuroinformatics is a fascinating research field that applies computational models and analytical tools to high dimensional experimental neuroscience data for a better understanding of how the brain functions or dysfunctions in brain diseases. Neuroinformaticians work in the intersection of neuroscience and informatics supporting the integration of various sub-disciplines (behavioural neuroscience, genetics, cognitive psychology, etc.) working on brain research. Neuroinformaticians are the pathway of information exchange between informaticians and clinicians for a better understanding of the outcome of computational models and the clinical interpretation of the analysis. Machine learning is one of the most significant computational developments in the last decade giving tools to neuroinformaticians and finally to radiologists and clinicians for an automatic and early diagnosis-prognosis of a brain disease. Random forest (RF) algorithm has been successfully applied to high-dimensional neuroimaging data for feature reduction and also has been applied to classify the clinical label of a subject using single or multi-modal neuroimaging datasets. Our aim was to review the studies where RF was applied to correctly predict the Alzheimer's disease (AD), the conversion from mild cognitive impairment (MCI) and its robustness to overfitting, outliers and handling of non-linear data. Finally, we described our RF-based model that gave us the 1 st position in an international challenge for automated prediction of MCI from MRI data.

Prediction of aquatic toxicity mode of action using linear discriminant and random forest models.

PubMed

Martin, Todd M; Grulke, Christopher M; Young, Douglas M; Russom, Christine L; Wang, Nina Y; Jackson, Crystal R; Barron, Mace G

2013-09-23

The ability to determine the mode of action (MOA) for a diverse group of chemicals is a critical part of ecological risk assessment and chemical regulation. However, existing MOA assignment approaches in ecotoxicology have been limited to a relatively few MOAs, have high uncertainty, or rely on professional judgment. In this study, machine based learning algorithms (linear discriminant analysis and random forest) were used to develop models for assigning aquatic toxicity MOA. These methods were selected since they have been shown to be able to correlate diverse data sets and provide an indication of the most important descriptors. A data set of MOA assignments for 924 chemicals was developed using a combination of high confidence assignments, international consensus classifications, ASTER (ASessment Tools for the Evaluation of Risk) predictions, and weight of evidence professional judgment based an assessment of structure and literature information. The overall data set was randomly divided into a training set (75%) and a validation set (25%) and then used to develop linear discriminant analysis (LDA) and random forest (RF) MOA assignment models. The LDA and RF models had high internal concordance and specificity and were able to produce overall prediction accuracies ranging from 84.5 to 87.7% for the validation set. These results demonstrate that computational chemistry approaches can be used to determine the acute toxicity MOAs across a large range of structures and mechanisms.

Random Forests (RFs) for Estimation, Uncertainty Prediction and Interpretation of Monthly Solar Potential

NASA Astrophysics Data System (ADS)

Assouline, Dan; Mohajeri, Nahid; Scartezzini, Jean-Louis

2017-04-01

Solar energy is clean, widely available, and arguably the most promising renewable energy resource. Taking full advantage of solar power, however, requires a deep understanding of its patterns and dependencies in space and time. The recent advances in Machine Learning brought powerful algorithms to estimate the spatio-temporal variations of solar irradiance (the power per unit area received from the Sun, W/m2), using local weather and terrain information. Such algorithms include Deep Learning (e.g. Artificial Neural Networks), or kernel methods (e.g. Support Vector Machines). However, most of these methods have some disadvantages, as they: (i) are complex to tune, (ii) are mainly used as a black box and offering no interpretation on the variables contributions, (iii) often do not provide uncertainty predictions (Assouline et al., 2016). To provide a reasonable solar mapping with good accuracy, these gaps would ideally need to be filled. We present here simple steps using one ensemble learning algorithm namely, Random Forests (Breiman, 2001) to (i) estimate monthly solar potential with good accuracy, (ii) provide information on the contribution of each feature in the estimation, and (iii) offer prediction intervals for each point estimate. We have selected Switzerland as an example. Using a Digital Elevation Model (DEM) along with monthly solar irradiance time series and weather data, we build monthly solar maps for Global Horizontal Irradiance (GHI), Diffuse Horizontal Irradiance (GHI), and Extraterrestrial Irradiance (EI). The weather data include monthly values for temperature, precipitation, sunshine duration, and cloud cover. In order to explain the impact of each feature on the solar irradiance of each point estimate, we extend the contribution method (Kuz'min et al., 2011) to a regression setting. Contribution maps for all features can then be computed for each solar map. This provides precious information on the spatial variation of the features impact all across Switzerland maps. Finally, as RFs are based on bootstrap samples of the training data, they can produce prediction intervals by looking at the trees estimates distribution, instead of taking the mean estimate. To do so, a simple idea is to grow all trees fully so that each leaf has exactly one value, that is, a training sample value. Then, for each point estimate, we compute percentiles of the trees estimates data to build a prediction interval. Two issues arise from this process: (i) growing the trees fully is not always possible, and (ii) there is a risk of over-fitting. We show how to solve them. These steps can be used for any type of environmental mapping so as to extract useful information on uncertainty and feature impact interpretation. References -Assouline, D., Mohajeri, N., & Scartezzini, J. L. (2017). Quantifying rooftop photovoltaic solar energy potential: A machine learning approach. Solar Energy, 141, 278-296. -Breiman, L. (2001). Random forests. Machine learning, 45(1), 5-32. -Kuz'min, V. E., Polishchuk, P. G., Artemenko, A. G., & Andronati, S. A. (2011). Interpretation of QSAR models based on random forest methods. Molecular Informatics, 30(6-7), 593-603.

Mathematical models application for mapping soils spatial distribution on the example of the farm from the North of Udmurt Republic of Russia

NASA Astrophysics Data System (ADS)

Dokuchaev, P. M.; Meshalkina, J. L.; Yaroslavtsev, A. M.

2018-01-01

Comparative analysis of soils geospatial modeling using multinomial logistic regression, decision trees, random forest, regression trees and support vector machines algorithms was conducted. The visual interpretation of the digital maps obtained and their comparison with the existing map, as well as the quantitative assessment of the individual soil groups detection overall accuracy and of the models kappa showed that multiple logistic regression, support vector method, and random forest models application with spatial prediction of the conditional soil groups distribution can be reliably used for mapping of the study area. It has shown the most accurate detection for sod-podzolics soils (Phaeozems Albic) lightly eroded and moderately eroded soils. In second place, according to the mean overall accuracy of the prediction, there are sod-podzolics soils - non-eroded and warp one, as well as sod-gley soils (Umbrisols Gleyic) and alluvial soils (Fluvisols Dystric, Umbric). Heavy eroded sod-podzolics and gray forest soils (Phaeozems Albic) were detected by methods of automatic classification worst of all.

Radio Frequency Interference Detection using Machine Learning.

NASA Astrophysics Data System (ADS)

Mosiane, Olorato; Oozeer, Nadeem; Aniyan, Arun; Bassett, Bruce A.

2017-05-01

Radio frequency interference (RFI) has plagued radio astronomy which potentially might be as bad or worse by the time the Square Kilometre Array (SKA) comes up. RFI can be either internal (generated by instruments) or external that originates from intentional or unintentional radio emission generated by man. With the huge amount of data that will be available with up coming radio telescopes, an automated aproach will be required to detect RFI. In this paper to try automate this process we present the result of applying machine learning techniques to cross match RFI from the Karoo Array Telescope (KAT-7) data. We found that not all the features selected to characterise RFI are always important. We further investigated 3 machine learning techniques and conclude that the Random forest classifier performs with a 98% Area Under Curve and 91% recall in detecting RFI.

Classifying bent radio galaxies from a mixture of point-like/extended images with Machine Learning.

NASA Astrophysics Data System (ADS)

Bastien, David; Oozeer, Nadeem; Somanah, Radhakrishna

2017-05-01

The hypothesis that bent radio sources are supposed to be found in rich, massive galaxy clusters and the avalibility of huge amount of data from radio surveys have fueled our motivation to use Machine Learning (ML) to identify bent radio sources and as such use them as tracers for galaxy clusters. The shapelet analysis allowed us to decompose radio images into 256 features that could be fed into the ML algorithm. Additionally, ideas from the field of neuro-psychology helped us to consider training the machine to identify bent galaxies at different orientations. From our analysis, we found that the Random Forest algorithm was the most effective with an accuracy rate of 92% for a classification of point and extended sources as well as an accuracy of 80% for bent and unbent classification.

Benchmarking dairy herd health status using routinely recorded herd summary data.

PubMed

Parker Gaddis, K L; Cole, J B; Clay, J S; Maltecca, C

2016-02-01

Genetic improvement of dairy cattle health through the use of producer-recorded data has been determined to be feasible. Low estimated heritabilities indicate that genetic progress will be slow. Variation observed in lowly heritable traits can largely be attributed to nongenetic factors, such as the environment. More rapid improvement of dairy cattle health may be attainable if herd health programs incorporate environmental and managerial aspects. More than 1,100 herd characteristics are regularly recorded on farm test-days. We combined these data with producer-recorded health event data, and parametric and nonparametric models were used to benchmark herd and cow health status. Health events were grouped into 3 categories for analyses: mastitis, reproductive, and metabolic. Both herd incidence and individual incidence were used as dependent variables. Models implemented included stepwise logistic regression, support vector machines, and random forests. At both the herd and individual levels, random forest models attained the highest accuracy for predicting health status in all health event categories when evaluated with 10-fold cross-validation. Accuracy (SD) ranged from 0.61 (0.04) to 0.63 (0.04) when using random forest models at the herd level. Accuracy of prediction (SD) at the individual cow level ranged from 0.87 (0.06) to 0.93 (0.001) with random forest models. Highly significant variables and key words from logistic regression and random forest models were also investigated. All models identified several of the same key factors for each health event category, including movement out of the herd, size of the herd, and weather-related variables. We concluded that benchmarking health status using routinely collected herd data is feasible. Nonparametric models were better suited to handle this complex data with numerous variables. These data mining techniques were able to perform prediction of health status and could add evidence to personal experience in herd management. Copyright © 2016 American Dairy Science Association. Published by Elsevier Inc. All rights reserved.

A Comparative Study with RapidMiner and WEKA Tools over some Classification Techniques for SMS Spam

NASA Astrophysics Data System (ADS)

Foozy, Cik Feresa Mohd; Ahmad, Rabiah; Faizal Abdollah, M. A.; Chai Wen, Chuah

2017-08-01

SMS Spamming is a serious attack that can manipulate the use of the SMS by spreading the advertisement in bulk. By sending the unwanted SMS that contain advertisement can make the users feeling disturb and this against the privacy of the mobile users. To overcome these issues, many studies have proposed to detect SMS Spam by using data mining tools. This paper will do a comparative study using five machine learning techniques such as Naïve Bayes, K-NN (K-Nearest Neighbour Algorithm), Decision Tree, Random Forest and Decision Stumps to observe the accuracy result between RapidMiner and WEKA for dataset SMS Spam UCI Machine Learning repository.

Machine-Learning Techniques for the Determination of Attrition of Forces Due to Atmospheric Conditions

DTIC Science & Technology

2018-02-01

the possibility of a correlation between aircraft incidents in the National Transportation Safety Board database and meteorological conditions. If a...strong correlation could be found, it could be used to derive a model to predict aircraft incidents and become part of a decision support tool for...techniques, primarily the random forest algorithm, were used to explore the possibility of a correlation between aircraft incidents in the National

Prediction of In-hospital Mortality in Emergency Department Patients With Sepsis: A Local Big Data-Driven, Machine Learning Approach.

PubMed

Taylor, R Andrew; Pare, Joseph R; Venkatesh, Arjun K; Mowafi, Hani; Melnick, Edward R; Fleischman, William; Hall, M Kennedy

2016-03-01

Predictive analytics in emergency care has mostly been limited to the use of clinical decision rules (CDRs) in the form of simple heuristics and scoring systems. In the development of CDRs, limitations in analytic methods and concerns with usability have generally constrained models to a preselected small set of variables judged to be clinically relevant and to rules that are easily calculated. Furthermore, CDRs frequently suffer from questions of generalizability, take years to develop, and lack the ability to be updated as new information becomes available. Newer analytic and machine learning techniques capable of harnessing the large number of variables that are already available through electronic health records (EHRs) may better predict patient outcomes and facilitate automation and deployment within clinical decision support systems. In this proof-of-concept study, a local, big data-driven, machine learning approach is compared to existing CDRs and traditional analytic methods using the prediction of sepsis in-hospital mortality as the use case. This was a retrospective study of adult ED visits admitted to the hospital meeting criteria for sepsis from October 2013 to October 2014. Sepsis was defined as meeting criteria for systemic inflammatory response syndrome with an infectious admitting diagnosis in the ED. ED visits were randomly partitioned into an 80%/20% split for training and validation. A random forest model (machine learning approach) was constructed using over 500 clinical variables from data available within the EHRs of four hospitals to predict in-hospital mortality. The machine learning prediction model was then compared to a classification and regression tree (CART) model, logistic regression model, and previously developed prediction tools on the validation data set using area under the receiver operating characteristic curve (AUC) and chi-square statistics. There were 5,278 visits among 4,676 unique patients who met criteria for sepsis. Of the 4,222 patients in the training group, 210 (5.0%) died during hospitalization, and of the 1,056 patients in the validation group, 50 (4.7%) died during hospitalization. The AUCs with 95% confidence intervals (CIs) for the different models were as follows: random forest model, 0.86 (95% CI = 0.82 to 0.90); CART model, 0.69 (95% CI = 0.62 to 0.77); logistic regression model, 0.76 (95% CI = 0.69 to 0.82); CURB-65, 0.73 (95% CI = 0.67 to 0.80); MEDS, 0.71 (95% CI = 0.63 to 0.77); and mREMS, 0.72 (95% CI = 0.65 to 0.79). The random forest model AUC was statistically different from all other models (p ≤ 0.003 for all comparisons). In this proof-of-concept study, a local big data-driven, machine learning approach outperformed existing CDRs as well as traditional analytic techniques for predicting in-hospital mortality of ED patients with sepsis. Future research should prospectively evaluate the effectiveness of this approach and whether it translates into improved clinical outcomes for high-risk sepsis patients. The methods developed serve as an example of a new model for predictive analytics in emergency care that can be automated, applied to other clinical outcomes of interest, and deployed in EHRs to enable locally relevant clinical predictions. © 2015 by the Society for Academic Emergency Medicine.

Prediction of In-hospital Mortality in Emergency Department Patients With Sepsis: A Local Big Data–Driven, Machine Learning Approach

PubMed Central

Taylor, R. Andrew; Pare, Joseph R.; Venkatesh, Arjun K.; Mowafi, Hani; Melnick, Edward R.; Fleischman, William; Hall, M. Kennedy

2018-01-01

Objectives Predictive analytics in emergency care has mostly been limited to the use of clinical decision rules (CDRs) in the form of simple heuristics and scoring systems. In the development of CDRs, limitations in analytic methods and concerns with usability have generally constrained models to a preselected small set of variables judged to be clinically relevant and to rules that are easily calculated. Furthermore, CDRs frequently suffer from questions of generalizability, take years to develop, and lack the ability to be updated as new information becomes available. Newer analytic and machine learning techniques capable of harnessing the large number of variables that are already available through electronic health records (EHRs) may better predict patient outcomes and facilitate automation and deployment within clinical decision support systems. In this proof-of-concept study, a local, big data–driven, machine learning approach is compared to existing CDRs and traditional analytic methods using the prediction of sepsis in-hospital mortality as the use case. Methods This was a retrospective study of adult ED visits admitted to the hospital meeting criteria for sepsis from October 2013 to October 2014. Sepsis was defined as meeting criteria for systemic inflammatory response syndrome with an infectious admitting diagnosis in the ED. ED visits were randomly partitioned into an 80%/20% split for training and validation. A random forest model (machine learning approach) was constructed using over 500 clinical variables from data available within the EHRs of four hospitals to predict in-hospital mortality. The machine learning prediction model was then compared to a classification and regression tree (CART) model, logistic regression model, and previously developed prediction tools on the validation data set using area under the receiver operating characteristic curve (AUC) and chi-square statistics. Results There were 5,278 visits among 4,676 unique patients who met criteria for sepsis. Of the 4,222 patients in the training group, 210 (5.0%) died during hospitalization, and of the 1,056 patients in the validation group, 50 (4.7%) died during hospitalization. The AUCs with 95% confidence intervals (CIs) for the different models were as follows: random forest model, 0.86 (95% CI = 0.82 to 0.90); CART model, 0.69 (95% CI = 0.62 to 0.77); logistic regression model, 0.76 (95% CI = 0.69 to 0.82); CURB-65, 0.73 (95% CI = 0.67 to 0.80); MEDS, 0.71 (95% CI = 0.63 to 0.77); and mREMS, 0.72 (95% CI = 0.65 to 0.79). The random forest model AUC was statistically different from all other models (p ≤ 0.003 for all comparisons). Conclusions In this proof-of-concept study, a local big data–driven, machine learning approach outperformed existing CDRs as well as traditional analytic techniques for predicting in-hospital mortality of ED patients with sepsis. Future research should prospectively evaluate the effectiveness of this approach and whether it translates into improved clinical outcomes for high-risk sepsis patients. The methods developed serve as an example of a new model for predictive analytics in emergency care that can be automated, applied to other clinical outcomes of interest, and deployed in EHRs to enable locally relevant clinical predictions. PMID:26679719

Building a profile of subjective well-being for social media users.

PubMed

Chen, Lushi; Gong, Tao; Kosinski, Michal; Stillwell, David; Davidson, Robert L

2017-01-01

Subjective well-being includes 'affect' and 'satisfaction with life' (SWL). This study proposes a unified approach to construct a profile of subjective well-being based on social media language in Facebook status updates. We apply sentiment analysis to generate users' affect scores, and train a random forest model to predict SWL using affect scores and other language features of the status updates. Results show that: the computer-selected features resemble the key predictors of SWL as identified in early studies; the machine-predicted SWL is moderately correlated with the self-reported SWL (r = 0.36, p < 0.01), indicating that language-based assessment can constitute valid SWL measures; the machine-assessed affect scores resemble those reported in a previous experimental study; and the machine-predicted subjective well-being profile can also reflect other psychological traits like depression (r = 0.24, p < 0.01). This study provides important insights for psychological prediction using multiple, machine-assessed components and longitudinal or dense psychological assessment using social media language.

Building a profile of subjective well-being for social media users

PubMed Central

Kosinski, Michal; Stillwell, David; Davidson, Robert L.

2017-01-01

Subjective well-being includes ‘affect’ and ‘satisfaction with life’ (SWL). This study proposes a unified approach to construct a profile of subjective well-being based on social media language in Facebook status updates. We apply sentiment analysis to generate users’ affect scores, and train a random forest model to predict SWL using affect scores and other language features of the status updates. Results show that: the computer-selected features resemble the key predictors of SWL as identified in early studies; the machine-predicted SWL is moderately correlated with the self-reported SWL (r = 0.36, p < 0.01), indicating that language-based assessment can constitute valid SWL measures; the machine-assessed affect scores resemble those reported in a previous experimental study; and the machine-predicted subjective well-being profile can also reflect other psychological traits like depression (r = 0.24, p < 0.01). This study provides important insights for psychological prediction using multiple, machine-assessed components and longitudinal or dense psychological assessment using social media language. PMID:29135991

Can We Train Machine Learning Methods to Outperform the High-dimensional Propensity Score Algorithm?

PubMed

Karim, Mohammad Ehsanul; Pang, Menglan; Platt, Robert W

2018-03-01

The use of retrospective health care claims datasets is frequently criticized for the lack of complete information on potential confounders. Utilizing patient's health status-related information from claims datasets as surrogates or proxies for mismeasured and unobserved confounders, the high-dimensional propensity score algorithm enables us to reduce bias. Using a previously published cohort study of postmyocardial infarction statin use (1998-2012), we compare the performance of the algorithm with a number of popular machine learning approaches for confounder selection in high-dimensional covariate spaces: random forest, least absolute shrinkage and selection operator, and elastic net. Our results suggest that, when the data analysis is done with epidemiologic principles in mind, machine learning methods perform as well as the high-dimensional propensity score algorithm. Using a plasmode framework that mimicked the empirical data, we also showed that a hybrid of machine learning and high-dimensional propensity score algorithms generally perform slightly better than both in terms of mean squared error, when a bias-based analysis is used.

Ensemble of random forests One vs. Rest classifiers for MCI and AD prediction using ANOVA cortical and subcortical feature selection and partial least squares.

PubMed

Ramírez, J; Górriz, J M; Ortiz, A; Martínez-Murcia, F J; Segovia, F; Salas-Gonzalez, D; Castillo-Barnes, D; Illán, I A; Puntonet, C G

2018-05-15

Alzheimer's disease (AD) is the most common cause of dementia in the elderly and affects approximately 30 million individuals worldwide. Mild cognitive impairment (MCI) is very frequently a prodromal phase of AD, and existing studies have suggested that people with MCI tend to progress to AD at a rate of about 10-15% per year. However, the ability of clinicians and machine learning systems to predict AD based on MRI biomarkers at an early stage is still a challenging problem that can have a great impact in improving treatments. The proposed system, developed by the SiPBA-UGR team for this challenge, is based on feature standardization, ANOVA feature selection, partial least squares feature dimension reduction and an ensemble of One vs. Rest random forest classifiers. With the aim of improving its performance when discriminating healthy controls (HC) from MCI, a second binary classification level was introduced that reconsiders the HC and MCI predictions of the first level. The system was trained and evaluated on an ADNI datasets that consist of T1-weighted MRI morphological measurements from HC, stable MCI, converter MCI and AD subjects. The proposed system yields a 56.25% classification score on the test subset which consists of 160 real subjects. The classifier yielded the best performance when compared to: (i) One vs. One (OvO), One vs. Rest (OvR) and error correcting output codes (ECOC) as strategies for reducing the multiclass classification task to multiple binary classification problems, (ii) support vector machines, gradient boosting classifier and random forest as base binary classifiers, and (iii) bagging ensemble learning. A robust method has been proposed for the international challenge on MCI prediction based on MRI data. The system yielded the second best performance during the competition with an accuracy rate of 56.25% when evaluated on the real subjects of the test set. Copyright © 2017 Elsevier B.V. All rights reserved.

A Stepwise Integrated Approach to Personalized Risk Predictions in Stage III Colorectal Cancer.

PubMed

Salvucci, Manuela; Würstle, Maximilian L; Morgan, Clare; Curry, Sarah; Cremona, Mattia; Lindner, Andreas U; Bacon, Orna; Resler, Alexa J; Murphy, Áine C; O'Byrne, Robert; Flanagan, Lorna; Dasgupta, Sonali; Rice, Nadege; Pilati, Camilla; Zink, Elisabeth; Schöller, Lisa M; Toomey, Sinead; Lawler, Mark; Johnston, Patrick G; Wilson, Richard; Camilleri-Broët, Sophie; Salto-Tellez, Manuel; McNamara, Deborah A; Kay, Elaine W; Laurent-Puig, Pierre; Van Schaeybroeck, Sandra; Hennessy, Bryan T; Longley, Daniel B; Rehm, Markus; Prehn, Jochen H M

2017-03-01

Purpose: Apoptosis is essential for chemotherapy responses. In this discovery and validation study, we evaluated the suitability of a mathematical model of apoptosis execution (APOPTO-CELL) as a stand-alone signature and as a constituent of further refined prognostic stratification tools. Experimental Design: Apoptosis competency of primary tumor samples from patients with stage III colorectal cancer ( n = 120) was calculated by APOPTO-CELL from measured protein concentrations of Procaspase-3, Procaspase-9, SMAC, and XIAP. An enriched APOPTO-CELL signature (APOPTO-CELL-PC3) was synthesized to capture apoptosome-independent effects of Caspase-3. Furthermore, a machine learning Random Forest approach was applied to APOPTO-CELL-PC3 and available molecular and clinicopathologic data to identify a further enhanced signature. Association of the signature with prognosis was evaluated in an independent colon adenocarcinoma cohort (TCGA COAD, n = 136). Results: We identified 3 prognostic biomarkers ( P = 0.04, P = 0.006, and P = 0.0004 for APOPTO-CELL, APOPTO-CELL-PC3, and Random Forest signatures, respectively) with increasing stratification accuracy for patients with stage III colorectal cancer.The APOPTO-CELL-PC3 signature ranked highest among all features. The prognostic value of the signatures was independently validated in stage III TCGA COAD patients ( P = 0.01, P = 0.04, and P = 0.02 for APOPTO-CELL, APOPTO-CELL-PC3, and Random Forest signatures, respectively). The signatures provided further stratification for patients with CMS1-3 molecular subtype. Conclusions: The integration of a systems-biology-based biomarker for apoptosis competency with machine learning approaches is an appealing and innovative strategy toward refined patient stratification. The prognostic value of apoptosis competency is independent of other available clinicopathologic and molecular factors, with tangible potential of being introduced in the clinical management of patients with stage III colorectal cancer. Clin Cancer Res; 23(5); 1200-12. ©2016 AACR . ©2016 American Association for Cancer Research.

Exposure assessment models for elemental components of particulate matter in an urban environment: A comparison of regression and random forest approaches

NASA Astrophysics Data System (ADS)

Brokamp, Cole; Jandarov, Roman; Rao, M. B.; LeMasters, Grace; Ryan, Patrick

2017-02-01

Exposure assessment for elemental components of particulate matter (PM) using land use modeling is a complex problem due to the high spatial and temporal variations in pollutant concentrations at the local scale. Land use regression (LUR) models may fail to capture complex interactions and non-linear relationships between pollutant concentrations and land use variables. The increasing availability of big spatial data and machine learning methods present an opportunity for improvement in PM exposure assessment models. In this manuscript, our objective was to develop a novel land use random forest (LURF) model and compare its accuracy and precision to a LUR model for elemental components of PM in the urban city of Cincinnati, Ohio. PM smaller than 2.5 μm (PM2.5) and eleven elemental components were measured at 24 sampling stations from the Cincinnati Childhood Allergy and Air Pollution Study (CCAAPS). Over 50 different predictors associated with transportation, physical features, community socioeconomic characteristics, greenspace, land cover, and emission point sources were used to construct LUR and LURF models. Cross validation was used to quantify and compare model performance. LURF and LUR models were created for aluminum (Al), copper (Cu), iron (Fe), potassium (K), manganese (Mn), nickel (Ni), lead (Pb), sulfur (S), silicon (Si), vanadium (V), zinc (Zn), and total PM2.5 in the CCAAPS study area. LURF utilized a more diverse and greater number of predictors than LUR and LURF models for Al, K, Mn, Pb, Si, Zn, TRAP, and PM2.5 all showed a decrease in fractional predictive error of at least 5% compared to their LUR models. LURF models for Al, Cu, Fe, K, Mn, Pb, Si, Zn, TRAP, and PM2.5 all had a cross validated fractional predictive error less than 30%. Furthermore, LUR models showed a differential exposure assessment bias and had a higher prediction error variance. Random forest and other machine learning methods may provide more accurate exposure assessment.

Exposure assessment models for elemental components of particulate matter in an urban environment: A comparison of regression and random forest approaches.

PubMed

Brokamp, Cole; Jandarov, Roman; Rao, M B; LeMasters, Grace; Ryan, Patrick

2017-02-01

Exposure assessment for elemental components of particulate matter (PM) using land use modeling is a complex problem due to the high spatial and temporal variations in pollutant concentrations at the local scale. Land use regression (LUR) models may fail to capture complex interactions and non-linear relationships between pollutant concentrations and land use variables. The increasing availability of big spatial data and machine learning methods present an opportunity for improvement in PM exposure assessment models. In this manuscript, our objective was to develop a novel land use random forest (LURF) model and compare its accuracy and precision to a LUR model for elemental components of PM in the urban city of Cincinnati, Ohio. PM smaller than 2.5 μm (PM2.5) and eleven elemental components were measured at 24 sampling stations from the Cincinnati Childhood Allergy and Air Pollution Study (CCAAPS). Over 50 different predictors associated with transportation, physical features, community socioeconomic characteristics, greenspace, land cover, and emission point sources were used to construct LUR and LURF models. Cross validation was used to quantify and compare model performance. LURF and LUR models were created for aluminum (Al), copper (Cu), iron (Fe), potassium (K), manganese (Mn), nickel (Ni), lead (Pb), sulfur (S), silicon (Si), vanadium (V), zinc (Zn), and total PM2.5 in the CCAAPS study area. LURF utilized a more diverse and greater number of predictors than LUR and LURF models for Al, K, Mn, Pb, Si, Zn, TRAP, and PM2.5 all showed a decrease in fractional predictive error of at least 5% compared to their LUR models. LURF models for Al, Cu, Fe, K, Mn, Pb, Si, Zn, TRAP, and PM2.5 all had a cross validated fractional predictive error less than 30%. Furthermore, LUR models showed a differential exposure assessment bias and had a higher prediction error variance. Random forest and other machine learning methods may provide more accurate exposure assessment.

Exposure assessment models for elemental components of particulate matter in an urban environment: A comparison of regression and random forest approaches

PubMed Central

Brokamp, Cole; Jandarov, Roman; Rao, M.B.; LeMasters, Grace; Ryan, Patrick

2017-01-01

Exposure assessment for elemental components of particulate matter (PM) using land use modeling is a complex problem due to the high spatial and temporal variations in pollutant concentrations at the local scale. Land use regression (LUR) models may fail to capture complex interactions and non-linear relationships between pollutant concentrations and land use variables. The increasing availability of big spatial data and machine learning methods present an opportunity for improvement in PM exposure assessment models. In this manuscript, our objective was to develop a novel land use random forest (LURF) model and compare its accuracy and precision to a LUR model for elemental components of PM in the urban city of Cincinnati, Ohio. PM smaller than 2.5 μm (PM2.5) and eleven elemental components were measured at 24 sampling stations from the Cincinnati Childhood Allergy and Air Pollution Study (CCAAPS). Over 50 different predictors associated with transportation, physical features, community socioeconomic characteristics, greenspace, land cover, and emission point sources were used to construct LUR and LURF models. Cross validation was used to quantify and compare model performance. LURF and LUR models were created for aluminum (Al), copper (Cu), iron (Fe), potassium (K), manganese (Mn), nickel (Ni), lead (Pb), sulfur (S), silicon (Si), vanadium (V), zinc (Zn), and total PM2.5 in the CCAAPS study area. LURF utilized a more diverse and greater number of predictors than LUR and LURF models for Al, K, Mn, Pb, Si, Zn, TRAP, and PM2.5 all showed a decrease in fractional predictive error of at least 5% compared to their LUR models. LURF models for Al, Cu, Fe, K, Mn, Pb, Si, Zn, TRAP, and PM2.5 all had a cross validated fractional predictive error less than 30%. Furthermore, LUR models showed a differential exposure assessment bias and had a higher prediction error variance. Random forest and other machine learning methods may provide more accurate exposure assessment. PMID:28959135

Novel approaches to assess the quality of fertility data stored in dairy herd management software.

PubMed

Hermans, K; Waegeman, W; Opsomer, G; Van Ranst, B; De Koster, J; Van Eetvelde, M; Hostens, M

2017-05-01

Scientific journals and popular press magazines are littered with articles in which the authors use data from dairy herd management software. Almost none of such papers include data cleaning and data quality assessment in their study design despite this being a very critical step during data mining. This paper presents 2 novel data cleaning methods that permit identification of animals with good and bad data quality. The first method is a deterministic or rule-based data cleaning method. Reproduction and mutation or life-changing events such as birth and death were converted to a symbolic (alphabetical letter) representation and split into triplets (3-letter code). The triplets were manually labeled as physiologically correct, suspicious, or impossible. The deterministic data cleaning method was applied to assess the quality of data stored in dairy herd management from 26 farms enrolled in the herd health management program from the Faculty of Veterinary Medicine Ghent University, Belgium. In total, 150,443 triplets were created, 65.4% were labeled as correct, 17.4% as suspicious, and 17.2% as impossible. The second method, a probabilistic method, uses a machine learning algorithm (random forests) to predict the correctness of fertility and mutation events in an early stage of data cleaning. The prediction accuracy of the random forests algorithm was compared with a classical linear statistical method (penalized logistic regression), outperforming the latter substantially, with a superior receiver operating characteristic curve and a higher accuracy (89 vs. 72%). From those results, we conclude that the triplet method can be used to assess the quality of reproduction data stored in dairy herd management software and that a machine learning technique such as random forests is capable of predicting the correctness of fertility data. Copyright © 2017 American Dairy Science Association. Published by Elsevier Inc. All rights reserved.

Minimizing soil impacts from forest operations

Treesearch

Emily A. Carter

2011-01-01

Several studies were conducted by Forest Service researchers and University and Industrial collaborators that investigated the potential for lessening soil surface disturbances and compaction in forest operations through modifications of machine components or harvest systems. Specific machine modifications included change in tire size, use of dual tire systems,...

Exploiting machine learning algorithms for tree species classification in a semiarid woodland using RapidEye image

NASA Astrophysics Data System (ADS)

Adelabu, Samuel; Mutanga, Onisimo; Adam, Elhadi; Cho, Moses Azong

2013-01-01

Classification of different tree species in semiarid areas can be challenging as a result of the change in leaf structure and orientation due to soil moisture constraints. Tree species mapping is, however, a key parameter for forest management in semiarid environments. In this study, we examined the suitability of 5-band RapidEye satellite data for the classification of five tree species in mopane woodland of Botswana using machine leaning algorithms with limited training samples.We performed classification using random forest (RF) and support vector machines (SVM) based on EnMap box. The overall accuracies for classifying the five tree species was 88.75 and 85% for both SVM and RF, respectively. We also demonstrated that the new red-edge band in the RapidEye sensor has the potential for classifying tree species in semiarid environments when integrated with other standard bands. Similarly, we observed that where there are limited training samples, SVM is preferred over RF. Finally, we demonstrated that the two accuracy measures of quantity and allocation disagreement are simpler and more helpful for the vast majority of remote sensing classification process than the kappa coefficient. Overall, high species classification can be achieved using strategically located RapidEye bands integrated with advanced processing algorithms.

Modeling Verdict Outcomes Using Social Network Measures: The Watergate and Caviar Network Cases

PubMed Central

2016-01-01

Modelling criminal trial verdict outcomes using social network measures is an emerging research area in quantitative criminology. Few studies have yet analyzed which of these measures are the most important for verdict modelling or which data classification techniques perform best for this application. To compare the performance of different techniques in classifying members of a criminal network, this article applies three different machine learning classifiers–Logistic Regression, Naïve Bayes and Random Forest–with a range of social network measures and the necessary databases to model the verdicts in two real–world cases: the U.S. Watergate Conspiracy of the 1970’s and the now–defunct Canada–based international drug trafficking ring known as the Caviar Network. In both cases it was found that the Random Forest classifier did better than either Logistic Regression or Naïve Bayes, and its superior performance was statistically significant. This being so, Random Forest was used not only for classification but also to assess the importance of the measures. For the Watergate case, the most important one proved to be betweenness centrality while for the Caviar Network, it was the effective size of the network. These results are significant because they show that an approach combining machine learning with social network analysis not only can generate accurate classification models but also helps quantify the importance social network variables in modelling verdict outcomes. We conclude our analysis with a discussion and some suggestions for future work in verdict modelling using social network measures. PMID:26824351

Estimating the domain of applicability for machine learning QSAR models: a study on aqueous solubility of drug discovery molecules.

PubMed

Schroeter, Timon Sebastian; Schwaighofer, Anton; Mika, Sebastian; Ter Laak, Antonius; Suelzle, Detlev; Ganzer, Ursula; Heinrich, Nikolaus; Müller, Klaus-Robert

2007-12-01

We investigate the use of different Machine Learning methods to construct models for aqueous solubility. Models are based on about 4000 compounds, including an in-house set of 632 drug discovery molecules of Bayer Schering Pharma. For each method, we also consider an appropriate method to obtain error bars, in order to estimate the domain of applicability (DOA) for each model. Here, we investigate error bars from a Bayesian model (Gaussian Process (GP)), an ensemble based approach (Random Forest), and approaches based on the Mahalanobis distance to training data (for Support Vector Machine and Ridge Regression models). We evaluate all approaches in terms of their prediction accuracy (in cross-validation, and on an external validation set of 536 molecules) and in how far the individual error bars can faithfully represent the actual prediction error.

Estimating the domain of applicability for machine learning QSAR models: a study on aqueous solubility of drug discovery molecules.

PubMed

Schroeter, Timon Sebastian; Schwaighofer, Anton; Mika, Sebastian; Ter Laak, Antonius; Suelzle, Detlev; Ganzer, Ursula; Heinrich, Nikolaus; Müller, Klaus-Robert

2007-09-01

We investigate the use of different Machine Learning methods to construct models for aqueous solubility. Models are based on about 4000 compounds, including an in-house set of 632 drug discovery molecules of Bayer Schering Pharma. For each method, we also consider an appropriate method to obtain error bars, in order to estimate the domain of applicability (DOA) for each model. Here, we investigate error bars from a Bayesian model (Gaussian Process (GP)), an ensemble based approach (Random Forest), and approaches based on the Mahalanobis distance to training data (for Support Vector Machine and Ridge Regression models). We evaluate all approaches in terms of their prediction accuracy (in cross-validation, and on an external validation set of 536 molecules) and in how far the individual error bars can faithfully represent the actual prediction error.

Estimating the domain of applicability for machine learning QSAR models: a study on aqueous solubility of drug discovery molecules

NASA Astrophysics Data System (ADS)

Schroeter, Timon Sebastian; Schwaighofer, Anton; Mika, Sebastian; Ter Laak, Antonius; Suelzle, Detlev; Ganzer, Ursula; Heinrich, Nikolaus; Müller, Klaus-Robert

2007-12-01

We investigate the use of different Machine Learning methods to construct models for aqueous solubility. Models are based on about 4000 compounds, including an in-house set of 632 drug discovery molecules of Bayer Schering Pharma. For each method, we also consider an appropriate method to obtain error bars, in order to estimate the domain of applicability (DOA) for each model. Here, we investigate error bars from a Bayesian model (Gaussian Process (GP)), an ensemble based approach (Random Forest), and approaches based on the Mahalanobis distance to training data (for Support Vector Machine and Ridge Regression models). We evaluate all approaches in terms of their prediction accuracy (in cross-validation, and on an external validation set of 536 molecules) and in how far the individual error bars can faithfully represent the actual prediction error.

Estimating the domain of applicability for machine learning QSAR models: a study on aqueous solubility of drug discovery molecules

NASA Astrophysics Data System (ADS)

Schroeter, Timon Sebastian; Schwaighofer, Anton; Mika, Sebastian; Ter Laak, Antonius; Suelzle, Detlev; Ganzer, Ursula; Heinrich, Nikolaus; Müller, Klaus-Robert

2007-09-01

We investigate the use of different Machine Learning methods to construct models for aqueous solubility. Models are based on about 4000 compounds, including an in-house set of 632 drug discovery molecules of Bayer Schering Pharma. For each method, we also consider an appropriate method to obtain error bars, in order to estimate the domain of applicability (DOA) for each model. Here, we investigate error bars from a Bayesian model (Gaussian Process (GP)), an ensemble based approach (Random Forest), and approaches based on the Mahalanobis distance to training data (for Support Vector Machine and Ridge Regression models). We evaluate all approaches in terms of their prediction accuracy (in cross-validation, and on an external validation set of 536 molecules) and in how far the individual error bars can faithfully represent the actual prediction error.

Geographical traceability of Marsdenia tenacissima by Fourier transform infrared spectroscopy and chemometrics

NASA Astrophysics Data System (ADS)

Li, Chao; Yang, Sheng-Chao; Guo, Qiao-Sheng; Zheng, Kai-Yan; Wang, Ping-Li; Meng, Zhen-Gui

2016-01-01

A combination of Fourier transform infrared spectroscopy with chemometrics tools provided an approach for studying Marsdenia tenacissima according to its geographical origin. A total of 128 M. tenacissima samples from four provinces in China were analyzed with FTIR spectroscopy. Six pattern recognition methods were used to construct the discrimination models: support vector machine-genetic algorithms, support vector machine-particle swarm optimization, K-nearest neighbors, radial basis function neural network, random forest and support vector machine-grid search. Experimental results showed that K-nearest neighbors was superior to other mathematical algorithms after data were preprocessed with wavelet de-noising, with a discrimination rate of 100% in both the training and prediction sets. This study demonstrated that FTIR spectroscopy coupled with K-nearest neighbors could be successfully applied to determine the geographical origins of M. tenacissima samples, thereby providing reliable authentication in a rapid, cheap and noninvasive way.

Predicting Survival From Large Echocardiography and Electronic Health Record Datasets: Optimization With Machine Learning.

PubMed

Samad, Manar D; Ulloa, Alvaro; Wehner, Gregory J; Jing, Linyuan; Hartzel, Dustin; Good, Christopher W; Williams, Brent A; Haggerty, Christopher M; Fornwalt, Brandon K

2018-06-09

The goal of this study was to use machine learning to more accurately predict survival after echocardiography. Predicting patient outcomes (e.g., survival) following echocardiography is primarily based on ejection fraction (EF) and comorbidities. However, there may be significant predictive information within additional echocardiography-derived measurements combined with clinical electronic health record data. Mortality was studied in 171,510 unselected patients who underwent 331,317 echocardiograms in a large regional health system. We investigated the predictive performance of nonlinear machine learning models compared with that of linear logistic regression models using 3 different inputs: 1) clinical variables, including 90 cardiovascular-relevant International Classification of Diseases, Tenth Revision, codes, and age, sex, height, weight, heart rate, blood pressures, low-density lipoprotein, high-density lipoprotein, and smoking; 2) clinical variables plus physician-reported EF; and 3) clinical variables and EF, plus 57 additional echocardiographic measurements. Missing data were imputed with a multivariate imputation by using a chained equations algorithm (MICE). We compared models versus each other and baseline clinical scoring systems by using a mean area under the curve (AUC) over 10 cross-validation folds and across 10 survival durations (6 to 60 months). Machine learning models achieved significantly higher prediction accuracy (all AUC >0.82) over common clinical risk scores (AUC = 0.61 to 0.79), with the nonlinear random forest models outperforming logistic regression (p < 0.01). The random forest model including all echocardiographic measurements yielded the highest prediction accuracy (p < 0.01 across all models and survival durations). Only 10 variables were needed to achieve 96% of the maximum prediction accuracy, with 6 of these variables being derived from echocardiography. Tricuspid regurgitation velocity was more predictive of survival than LVEF. In a subset of studies with complete data for the top 10 variables, multivariate imputation by chained equations yielded slightly reduced predictive accuracies (difference in AUC of 0.003) compared with the original data. Machine learning can fully utilize large combinations of disparate input variables to predict survival after echocardiography with superior accuracy. Copyright © 2018 American College of Cardiology Foundation. Published by Elsevier Inc. All rights reserved.

Defining Higher-Order Turbulent Moment Closures with an Artificial Neural Network and Random Forest

NASA Astrophysics Data System (ADS)

McGibbon, J.; Bretherton, C. S.

2017-12-01

Unresolved turbulent advection and clouds must be parameterized in atmospheric models. Modern higher-order closure schemes depend on analytic moment closure assumptions that diagnose higher-order moments in terms of lower-order ones. These are then tested against Large-Eddy Simulation (LES) higher-order moment relations. However, these relations may not be neatly analytic in nature. Rather than rely on an analytic higher-order moment closure, can we use machine learning on LES data itself to define a higher-order moment closure?We assess the ability of a deep artificial neural network (NN) and random forest (RF) to perform this task using a set of observationally-based LES runs from the MAGIC field campaign. By training on a subset of 12 simulations and testing on remaining simulations, we avoid over-fitting the training data.Performance of the NN and RF will be assessed and compared to the Analytic Double Gaussian 1 (ADG1) closure assumed by Cloudy Layers Unified By Binormals (CLUBB), a higher-order turbulence closure currently used in the Community Atmosphere Model (CAM). We will show that the RF outperforms the NN and the ADG1 closure for the MAGIC cases within this diagnostic framework. Progress and challenges in using a diagnostic machine learning closure within a prognostic cloud and turbulence parameterization will also be discussed.

Lawsuit lead time prediction: Comparison of data mining techniques based on categorical response variable.

PubMed

Gruginskie, Lúcia Adriana Dos Santos; Vaccaro, Guilherme Luís Roehe

2018-01-01

The quality of the judicial system of a country can be verified by the overall length time of lawsuits, or the lead time. When the lead time is excessive, a country's economy can be affected, leading to the adoption of measures such as the creation of the Saturn Center in Europe. Although there are performance indicators to measure the lead time of lawsuits, the analysis and the fit of prediction models are still underdeveloped themes in the literature. To contribute to this subject, this article compares different prediction models according to their accuracy, sensitivity, specificity, precision, and F1 measure. The database used was from TRF4-the Tribunal Regional Federal da 4a Região-a federal court in southern Brazil, corresponding to the 2nd Instance civil lawsuits completed in 2016. The models were fitted using support vector machine, naive Bayes, random forests, and neural network approaches with categorical predictor variables. The lead time of the 2nd Instance judgment was selected as the response variable measured in days and categorized in bands. The comparison among the models showed that the support vector machine and random forest approaches produced measurements that were superior to those of the other models. The evaluation of the models was made using k-fold cross-validation similar to that applied to the test models.

Mirnacle: machine learning with SMOTE and random forest for improving selectivity in pre-miRNA ab initio prediction.

PubMed

Marques, Yuri Bento; de Paiva Oliveira, Alcione; Ribeiro Vasconcelos, Ana Tereza; Cerqueira, Fabio Ribeiro

2016-12-15

MicroRNAs (miRNAs) are key gene expression regulators in plants and animals. Therefore, miRNAs are involved in several biological processes, making the study of these molecules one of the most relevant topics of molecular biology nowadays. However, characterizing miRNAs in vivo is still a complex task. As a consequence, in silico methods have been developed to predict miRNA loci. A common ab initio strategy to find miRNAs in genomic data is to search for sequences that can fold into the typical hairpin structure of miRNA precursors (pre-miRNAs). The current ab initio approaches, however, have selectivity issues, i.e., a high number of false positives is reported, which can lead to laborious and costly attempts to provide biological validation. This study presents an extension of the ab initio method miRNAFold, with the aim of improving selectivity through machine learning techniques, namely, random forest combined with the SMOTE procedure that copes with imbalance datasets. By comparing our method, termed Mirnacle, with other important approaches in the literature, we demonstrate that Mirnacle substantially improves selectivity without compromising sensitivity. For the three datasets used in our experiments, our method achieved at least 97% of sensitivity and could deliver a two-fold, 20-fold, and 6-fold increase in selectivity, respectively, compared with the best results of current computational tools. The extension of miRNAFold by the introduction of machine learning techniques, significantly increases selectivity in pre-miRNA ab initio prediction, which optimally contributes to advanced studies on miRNAs, as the need of biological validations is diminished. Hopefully, new research, such as studies of severe diseases caused by miRNA malfunction, will benefit from the proposed computational tool.

Automatic detection of anatomical regions in frontal x-ray images: comparing convolutional neural networks to random forest

NASA Astrophysics Data System (ADS)

Olory Agomma, R.; Vázquez, C.; Cresson, T.; De Guise, J.

2018-02-01

Most algorithms to detect and identify anatomical structures in medical images require either to be initialized close to the target structure, or to know that the structure is present in the image, or to be trained on a homogeneous database (e.g. all full body or all lower limbs). Detecting these structures when there is no guarantee that the structure is present in the image, or when the image database is heterogeneous (mixed configurations), is a challenge for automatic algorithms. In this work we compared two state-of-the-art machine learning techniques in order to determine which one is the most appropriate for predicting targets locations based on image patches. By knowing the position of thirteen landmarks points, labelled by an expert in EOS frontal radiography, we learn the displacement between salient points detected in the image and these thirteen landmarks. The learning step is carried out with a machine learning approach by exploring two methods: Convolutional Neural Network (CNN) and Random Forest (RF). The automatic detection of the thirteen landmarks points in a new image is then obtained by averaging the positions of each one of these thirteen landmarks estimated from all the salient points in the new image. We respectively obtain for CNN and RF, an average prediction error (both mean and standard deviation in mm) of 29 +/-18 and 30 +/- 21 for the thirteen landmarks points, indicating the approximate location of anatomical regions. On the other hand, the learning time is 9 days for CNN versus 80 minutes for RF. We provide a comparison of the results between the two machine learning approaches.

Combining deep residual neural network features with supervised machine learning algorithms to classify diverse food image datasets.

PubMed

McAllister, Patrick; Zheng, Huiru; Bond, Raymond; Moorhead, Anne

2018-04-01

Obesity is increasing worldwide and can cause many chronic conditions such as type-2 diabetes, heart disease, sleep apnea, and some cancers. Monitoring dietary intake through food logging is a key method to maintain a healthy lifestyle to prevent and manage obesity. Computer vision methods have been applied to food logging to automate image classification for monitoring dietary intake. In this work we applied pretrained ResNet-152 and GoogleNet convolutional neural networks (CNNs), initially trained using ImageNet Large Scale Visual Recognition Challenge (ILSVRC) dataset with MatConvNet package, to extract features from food image datasets; Food 5K, Food-11, RawFooT-DB, and Food-101. Deep features were extracted from CNNs and used to train machine learning classifiers including artificial neural network (ANN), support vector machine (SVM), Random Forest, and Naive Bayes. Results show that using ResNet-152 deep features with SVM with RBF kernel can accurately detect food items with 99.4% accuracy using Food-5K validation food image dataset and 98.8% with Food-5K evaluation dataset using ANN, SVM-RBF, and Random Forest classifiers. Trained with ResNet-152 features, ANN can achieve 91.34%, 99.28% when applied to Food-11 and RawFooT-DB food image datasets respectively and SVM with RBF kernel can achieve 64.98% with Food-101 image dataset. From this research it is clear that using deep CNN features can be used efficiently for diverse food item image classification. The work presented in this research shows that pretrained ResNet-152 features provide sufficient generalisation power when applied to a range of food image classification tasks. Copyright © 2018 Elsevier Ltd. All rights reserved.

Identification of immune correlates of protection in Shigella infection by application of machine learning.

PubMed

Arevalillo, Jorge M; Sztein, Marcelo B; Kotloff, Karen L; Levine, Myron M; Simon, Jakub K

2017-10-01

Immunologic correlates of protection are important in vaccine development because they give insight into mechanisms of protection, assist in the identification of promising vaccine candidates, and serve as endpoints in bridging clinical vaccine studies. Our goal is the development of a methodology to identify immunologic correlates of protection using the Shigella challenge as a model. The proposed methodology utilizes the Random Forests (RF) machine learning algorithm as well as Classification and Regression Trees (CART) to detect immune markers that predict protection, identify interactions between variables, and define optimal cutoffs. Logistic regression modeling is applied to estimate the probability of protection and the confidence interval (CI) for such a probability is computed by bootstrapping the logistic regression models. The results demonstrate that the combination of Classification and Regression Trees and Random Forests complements the standard logistic regression and uncovers subtle immune interactions. Specific levels of immunoglobulin IgG antibody in blood on the day of challenge predicted protection in 75% (95% CI 67-86). Of those subjects that did not have blood IgG at or above a defined threshold, 100% were protected if they had IgA antibody secreting cells above a defined threshold. Comparison with the results obtained by applying only logistic regression modeling with standard Akaike Information Criterion for model selection shows the usefulness of the proposed method. Given the complexity of the immune system, the use of machine learning methods may enhance traditional statistical approaches. When applied together, they offer a novel way to quantify important immune correlates of protection that may help the development of vaccines. Copyright © 2017 Elsevier Inc. All rights reserved.

Subtyping cognitive profiles in Autism Spectrum Disorder using a Functional Random Forest algorithm.

PubMed

Feczko, E; Balba, N M; Miranda-Dominguez, O; Cordova, M; Karalunas, S L; Irwin, L; Demeter, D V; Hill, A P; Langhorst, B H; Grieser Painter, J; Van Santen, J; Fombonne, E J; Nigg, J T; Fair, D A

2018-05-15

DSM-5 Autism Spectrum Disorder (ASD) comprises a set of neurodevelopmental disorders characterized by deficits in social communication and interaction and repetitive behaviors or restricted interests, and may both affect and be affected by multiple cognitive mechanisms. This study attempts to identify and characterize cognitive subtypes within the ASD population using our Functional Random Forest (FRF) machine learning classification model. This model trained a traditional random forest model on measures from seven tasks that reflect multiple levels of information processing. 47 ASD diagnosed and 58 typically developing (TD) children between the ages of 9 and 13 participated in this study. Our RF model was 72.7% accurate, with 80.7% specificity and 63.1% sensitivity. Using the random forest model, the FRF then measures the proximity of each subject to every other subject, generating a distance matrix between participants. This matrix is then used in a community detection algorithm to identify subgroups within the ASD and TD groups, and revealed 3 ASD and 4 TD putative subgroups with unique behavioral profiles. We then examined differences in functional brain systems between diagnostic groups and putative subgroups using resting-state functional connectivity magnetic resonance imaging (rsfcMRI). Chi-square tests revealed a significantly greater number of between group differences (p < .05) within the cingulo-opercular, visual, and default systems as well as differences in inter-system connections in the somato-motor, dorsal attention, and subcortical systems. Many of these differences were primarily driven by specific subgroups suggesting that our method could potentially parse the variation in brain mechanisms affected by ASD. Copyright © 2017. Published by Elsevier Inc.

Properties of Protein Drug Target Classes

PubMed Central

Bull, Simon C.; Doig, Andrew J.

2015-01-01

Accurate identification of drug targets is a crucial part of any drug development program. We mined the human proteome to discover properties of proteins that may be important in determining their suitability for pharmaceutical modulation. Data was gathered concerning each protein’s sequence, post-translational modifications, secondary structure, germline variants, expression profile and drug target status. The data was then analysed to determine features for which the target and non-target proteins had significantly different values. This analysis was repeated for subsets of the proteome consisting of all G-protein coupled receptors, ion channels, kinases and proteases, as well as proteins that are implicated in cancer. Machine learning was used to quantify the proteins in each dataset in terms of their potential to serve as a drug target. This was accomplished by first inducing a random forest that could distinguish between its targets and non-targets, and then using the random forest to quantify the drug target likeness of the non-targets. The properties that can best differentiate targets from non-targets were primarily those that are directly related to a protein’s sequence (e.g. secondary structure). Germline variants, expression levels and interactions between proteins had minimal discriminative power. Overall, the best indicators of drug target likeness were found to be the proteins’ hydrophobicities, in vivo half-lives, propensity for being membrane bound and the fraction of non-polar amino acids in their sequences. In terms of predicting potential targets, datasets of proteases, ion channels and cancer proteins were able to induce random forests that were highly capable of distinguishing between targets and non-targets. The non-target proteins predicted to be targets by these random forests comprise the set of the most suitable potential future drug targets, and should therefore be prioritised when building a drug development programme. PMID:25822509

Random Forest Algorithm for the Classification of Neuroimaging Data in Alzheimer's Disease: A Systematic Review.

PubMed

Sarica, Alessia; Cerasa, Antonio; Quattrone, Aldo

2017-01-01

Objective: Machine learning classification has been the most important computational development in the last years to satisfy the primary need of clinicians for automatic early diagnosis and prognosis. Nowadays, Random Forest (RF) algorithm has been successfully applied for reducing high dimensional and multi-source data in many scientific realms. Our aim was to explore the state of the art of the application of RF on single and multi-modal neuroimaging data for the prediction of Alzheimer's disease. Methods: A systematic review following PRISMA guidelines was conducted on this field of study. In particular, we constructed an advanced query using boolean operators as follows: ("random forest" OR "random forests") AND neuroimaging AND ("alzheimer's disease" OR alzheimer's OR alzheimer) AND (prediction OR classification) . The query was then searched in four well-known scientific databases: Pubmed, Scopus, Google Scholar and Web of Science. Results: Twelve articles-published between the 2007 and 2017-have been included in this systematic review after a quantitative and qualitative selection. The lesson learnt from these works suggest that when RF was applied on multi-modal data for prediction of Alzheimer's disease (AD) conversion from the Mild Cognitive Impairment (MCI), it produces one of the best accuracies to date. Moreover, the RF has important advantages in terms of robustness to overfitting, ability to handle highly non-linear data, stability in the presence of outliers and opportunity for efficient parallel processing mainly when applied on multi-modality neuroimaging data, such as, MRI morphometric, diffusion tensor imaging, and PET images. Conclusions: We discussed the strengths of RF, considering also possible limitations and by encouraging further studies on the comparisons of this algorithm with other commonly used classification approaches, particularly in the early prediction of the progression from MCI to AD.

Evolving optimised decision rules for intrusion detection using particle swarm paradigm

NASA Astrophysics Data System (ADS)

Sivatha Sindhu, Siva S.; Geetha, S.; Kannan, A.

2012-12-01

The aim of this article is to construct a practical intrusion detection system (IDS) that properly analyses the statistics of network traffic pattern and classify them as normal or anomalous class. The objective of this article is to prove that the choice of effective network traffic features and a proficient machine-learning paradigm enhances the detection accuracy of IDS. In this article, a rule-based approach with a family of six decision tree classifiers, namely Decision Stump, C4.5, Naive Baye's Tree, Random Forest, Random Tree and Representative Tree model to perform the detection of anomalous network pattern is introduced. In particular, the proposed swarm optimisation-based approach selects instances that compose training set and optimised decision tree operate over this trained set producing classification rules with improved coverage, classification capability and generalisation ability. Experiment with the Knowledge Discovery and Data mining (KDD) data set which have information on traffic pattern, during normal and intrusive behaviour shows that the proposed algorithm produces optimised decision rules and outperforms other machine-learning algorithm.

Recognising discourse causality triggers in the biomedical domain.

PubMed

Mihăilă, Claudiu; Ananiadou, Sophia

2013-12-01

Current domain-specific information extraction systems represent an important resource for biomedical researchers, who need to process vast amounts of knowledge in a short time. Automatic discourse causality recognition can further reduce their workload by suggesting possible causal connections and aiding in the curation of pathway models. We describe here an approach to the automatic identification of discourse causality triggers in the biomedical domain using machine learning. We create several baselines and experiment with and compare various parameter settings for three algorithms, i.e. Conditional Random Fields (CRF), Support Vector Machines (SVM) and Random Forests (RF). We also evaluate the impact of lexical, syntactic, and semantic features on each of the algorithms, showing that semantics improves the performance in all cases. We test our comprehensive feature set on two corpora containing gold standard annotations of causal relations, and demonstrate the need for more gold standard data. The best performance of 79.35% F-score is achieved by CRFs when using all three feature types.

New developments in operator protection for forest machines

Treesearch

Robert B. Rummer; S. Taylor; M. Veal

2003-01-01

Mechanization of forest operations ha greatly improved saftey of woods work. However, increasing use of machines has introduced new hazards that must be addressed. Two of these hazards are rollover of swing-type forestry machines (currently excluded from standard protection) and the hazard of thrown objects from cutting devices. Ongoing research projects are developing...

METAPHOR: Probability density estimation for machine learning based photometric redshifts

NASA Astrophysics Data System (ADS)

Amaro, V.; Cavuoti, S.; Brescia, M.; Vellucci, C.; Tortora, C.; Longo, G.

2017-06-01

We present METAPHOR (Machine-learning Estimation Tool for Accurate PHOtometric Redshifts), a method able to provide a reliable PDF for photometric galaxy redshifts estimated through empirical techniques. METAPHOR is a modular workflow, mainly based on the MLPQNA neural network as internal engine to derive photometric galaxy redshifts, but giving the possibility to easily replace MLPQNA with any other method to predict photo-z's and their PDF. We present here the results about a validation test of the workflow on the galaxies from SDSS-DR9, showing also the universality of the method by replacing MLPQNA with KNN and Random Forest models. The validation test include also a comparison with the PDF's derived from a traditional SED template fitting method (Le Phare).

Feature-Free Activity Classification of Inertial Sensor Data With Machine Vision Techniques: Method, Development, and Evaluation.

PubMed

Dominguez Veiga, Jose Juan; O'Reilly, Martin; Whelan, Darragh; Caulfield, Brian; Ward, Tomas E

2017-08-04

Inertial sensors are one of the most commonly used sources of data for human activity recognition (HAR) and exercise detection (ED) tasks. The time series produced by these sensors are generally analyzed through numerical methods. Machine learning techniques such as random forests or support vector machines are popular in this field for classification efforts, but they need to be supported through the isolation of a potentially large number of additionally crafted features derived from the raw data. This feature preprocessing step can involve nontrivial digital signal processing (DSP) techniques. However, in many cases, the researchers interested in this type of activity recognition problems do not possess the necessary technical background for this feature-set development. The study aimed to present a novel application of established machine vision methods to provide interested researchers with an easier entry path into the HAR and ED fields. This can be achieved by removing the need for deep DSP skills through the use of transfer learning. This can be done by using a pretrained convolutional neural network (CNN) developed for machine vision purposes for exercise classification effort. The new method should simply require researchers to generate plots of the signals that they would like to build classifiers with, store them as images, and then place them in folders according to their training label before retraining the network. We applied a CNN, an established machine vision technique, to the task of ED. Tensorflow, a high-level framework for machine learning, was used to facilitate infrastructure needs. Simple time series plots generated directly from accelerometer and gyroscope signals are used to retrain an openly available neural network (Inception), originally developed for machine vision tasks. Data from 82 healthy volunteers, performing 5 different exercises while wearing a lumbar-worn inertial measurement unit (IMU), was collected. The ability of the proposed method to automatically classify the exercise being completed was assessed using this dataset. For comparative purposes, classification using the same dataset was also performed using the more conventional approach of feature-extraction and classification using random forest classifiers. With the collected dataset and the proposed method, the different exercises could be recognized with a 95.89% (3827/3991) accuracy, which is competitive with current state-of-the-art techniques in ED. The high level of accuracy attained with the proposed approach indicates that the waveform morphologies in the time-series plots for each of the exercises is sufficiently distinct among the participants to allow the use of machine vision approaches. The use of high-level machine learning frameworks, coupled with the novel use of machine vision techniques instead of complex manually crafted features, may facilitate access to research in the HAR field for individuals without extensive digital signal processing or machine learning backgrounds. ©Jose Juan Dominguez Veiga, Martin O'Reilly, Darragh Whelan, Brian Caulfield, Tomas E Ward. Originally published in JMIR Mhealth and Uhealth (http://mhealth.jmir.org), 04.08.2017.

Feature-Free Activity Classification of Inertial Sensor Data With Machine Vision Techniques: Method, Development, and Evaluation

PubMed Central

O'Reilly, Martin; Whelan, Darragh; Caulfield, Brian; Ward, Tomas E

2017-01-01

Background Inertial sensors are one of the most commonly used sources of data for human activity recognition (HAR) and exercise detection (ED) tasks. The time series produced by these sensors are generally analyzed through numerical methods. Machine learning techniques such as random forests or support vector machines are popular in this field for classification efforts, but they need to be supported through the isolation of a potentially large number of additionally crafted features derived from the raw data. This feature preprocessing step can involve nontrivial digital signal processing (DSP) techniques. However, in many cases, the researchers interested in this type of activity recognition problems do not possess the necessary technical background for this feature-set development. Objective The study aimed to present a novel application of established machine vision methods to provide interested researchers with an easier entry path into the HAR and ED fields. This can be achieved by removing the need for deep DSP skills through the use of transfer learning. This can be done by using a pretrained convolutional neural network (CNN) developed for machine vision purposes for exercise classification effort. The new method should simply require researchers to generate plots of the signals that they would like to build classifiers with, store them as images, and then place them in folders according to their training label before retraining the network. Methods We applied a CNN, an established machine vision technique, to the task of ED. Tensorflow, a high-level framework for machine learning, was used to facilitate infrastructure needs. Simple time series plots generated directly from accelerometer and gyroscope signals are used to retrain an openly available neural network (Inception), originally developed for machine vision tasks. Data from 82 healthy volunteers, performing 5 different exercises while wearing a lumbar-worn inertial measurement unit (IMU), was collected. The ability of the proposed method to automatically classify the exercise being completed was assessed using this dataset. For comparative purposes, classification using the same dataset was also performed using the more conventional approach of feature-extraction and classification using random forest classifiers. Results With the collected dataset and the proposed method, the different exercises could be recognized with a 95.89% (3827/3991) accuracy, which is competitive with current state-of-the-art techniques in ED. Conclusions The high level of accuracy attained with the proposed approach indicates that the waveform morphologies in the time-series plots for each of the exercises is sufficiently distinct among the participants to allow the use of machine vision approaches. The use of high-level machine learning frameworks, coupled with the novel use of machine vision techniques instead of complex manually crafted features, may facilitate access to research in the HAR field for individuals without extensive digital signal processing or machine learning backgrounds. PMID:28778851

DOE Office of Scientific and Technical Information (OSTI.GOV)

Ukwatta, T. N.; Wozniak, P. R.; Gehrels, N.

Studies of high-redshift gamma-ray bursts (GRBs) provide important information about the early Universe such as the rates of stellar collapsars and mergers, the metallicity content, constraints on the re-ionization period, and probes of the Hubble expansion. Rapid selection of high-z candidates from GRB samples reported in real time by dedicated space missions such as Swift is the key to identifying the most distant bursts before the optical afterglow becomes too dim to warrant a good spectrum. Here, we introduce ‘machine-z’, a redshift prediction algorithm and a ‘high-z’ classifier for Swift GRBs based on machine learning. Our method relies exclusively onmore » canonical data commonly available within the first few hours after the GRB trigger. Using a sample of 284 bursts with measured redshifts, we trained a randomized ensemble of decision trees (random forest) to perform both regression and classification. Cross-validated performance studies show that the correlation coefficient between machine-z predictions and the true redshift is nearly 0.6. At the same time, our high-z classifier can achieve 80 per cent recall of true high-redshift bursts, while incurring a false positive rate of 20 per cent. With 40 per cent false positive rate the classifier can achieve ~100 per cent recall. As a result, the most reliable selection of high-redshift GRBs is obtained by combining predictions from both the high-z classifier and the machine-z regressor.« less

Thrown object testing of forest machine operator protective structures

Treesearch

S.E. Taylor; M.W. Veal; R.B. Rummer

2003-01-01

High-speed chains or rotating disks are commonly used to cut and process trees during forest harvesting operations. Mechanical failure or fatigue of these tools can lead to a potentially hazardous situation where fragments of chain or sawteeth are thrown through the operator enclosures on forest machines. This poster presentation discusses the development and...

Prediction of antiepileptic drug treatment outcomes using machine learning.

PubMed

Colic, Sinisa; Wither, Robert G; Lang, Min; Zhang, Liang; Eubanks, James H; Bardakjian, Berj L

2017-02-01

Antiepileptic drug (AED) treatments produce inconsistent outcomes, often necessitating patients to go through several drug trials until a successful treatment can be found. This study proposes the use of machine learning techniques to predict epilepsy treatment outcomes of commonly used AEDs. Machine learning algorithms were trained and evaluated using features obtained from intracranial electroencephalogram (iEEG) recordings of the epileptiform discharges observed in Mecp2-deficient mouse model of the Rett Syndrome. Previous work have linked the presence of cross-frequency coupling (I CFC ) of the delta (2-5 Hz) rhythm with the fast ripple (400-600 Hz) rhythm in epileptiform discharges. Using the I CFC to label post-treatment outcomes we compared support vector machines (SVMs) and random forest (RF) machine learning classifiers for providing likelihood scores of successful treatment outcomes. (a) There was heterogeneity in AED treatment outcomes, (b) machine learning techniques could be used to rank the efficacy of AEDs by estimating likelihood scores for successful treatment outcome, (c) I CFC features yielded the most effective a priori identification of appropriate AED treatment, and (d) both classifiers performed comparably. Machine learning approaches yielded predictions of successful drug treatment outcomes which in turn could reduce the burdens of drug trials and lead to substantial improvements in patient quality of life.

Prediction of antiepileptic drug treatment outcomes using machine learning

NASA Astrophysics Data System (ADS)

Colic, Sinisa; Wither, Robert G.; Lang, Min; Zhang, Liang; Eubanks, James H.; Bardakjian, Berj L.

2017-02-01

Objective. Antiepileptic drug (AED) treatments produce inconsistent outcomes, often necessitating patients to go through several drug trials until a successful treatment can be found. This study proposes the use of machine learning techniques to predict epilepsy treatment outcomes of commonly used AEDs. Approach. Machine learning algorithms were trained and evaluated using features obtained from intracranial electroencephalogram (iEEG) recordings of the epileptiform discharges observed in Mecp2-deficient mouse model of the Rett Syndrome. Previous work have linked the presence of cross-frequency coupling (I CFC) of the delta (2-5 Hz) rhythm with the fast ripple (400-600 Hz) rhythm in epileptiform discharges. Using the I CFC to label post-treatment outcomes we compared support vector machines (SVMs) and random forest (RF) machine learning classifiers for providing likelihood scores of successful treatment outcomes. Main results. (a) There was heterogeneity in AED treatment outcomes, (b) machine learning techniques could be used to rank the efficacy of AEDs by estimating likelihood scores for successful treatment outcome, (c) I CFC features yielded the most effective a priori identification of appropriate AED treatment, and (d) both classifiers performed comparably. Significance. Machine learning approaches yielded predictions of successful drug treatment outcomes which in turn could reduce the burdens of drug trials and lead to substantial improvements in patient quality of life.

Predicting Seagrass Occurrence in a Changing Climate Using Random Forests

NASA Astrophysics Data System (ADS)

Aydin, O.; Butler, K. A.

2017-12-01

Seagrasses are marine plants that can quickly sequester vast amounts of carbon (up to 100 times more and 12 times faster than tropical forests). In this work, we present an integrated GIS and machine learning approach to build a data-driven model of seagrass presence-absence. We outline a random forest approach that avoids the prevalence bias in many ecological presence-absence models. One of our goals is to predict global seagrass occurrence from a spatially limited training sample. In addition, we conduct a sensitivity study which investigates the vulnerability of seagrass to changing climate conditions. We integrate multiple data sources including fine-scale seagrass data from MarineCadastre.gov and the recently available globally extensive publicly available Ecological Marine Units (EMU) dataset. These data are used to train a model for seagrass occurrence along the U.S. coast. In situ oceans data are interpolated using Empirical Bayesian Kriging (EBK) to produce globally extensive prediction variables. A neural network is used to estimate probable future values of prediction variables such as ocean temperature to assess the impact of a warming climate on seagrass occurrence. The proposed workflow can be generalized to many presence-absence models.

Optimizing classification performance in an object-based very-high-resolution land use-land cover urban application

NASA Astrophysics Data System (ADS)

Georganos, Stefanos; Grippa, Tais; Vanhuysse, Sabine; Lennert, Moritz; Shimoni, Michal; Wolff, Eléonore

2017-10-01

This study evaluates the impact of three Feature Selection (FS) algorithms in an Object Based Image Analysis (OBIA) framework for Very-High-Resolution (VHR) Land Use-Land Cover (LULC) classification. The three selected FS algorithms, Correlation Based Selection (CFS), Mean Decrease in Accuracy (MDA) and Random Forest (RF) based Recursive Feature Elimination (RFE), were tested on Support Vector Machine (SVM), K-Nearest Neighbor, and Random Forest (RF) classifiers. The results demonstrate that the accuracy of SVM and KNN classifiers are the most sensitive to FS. The RF appeared to be more robust to high dimensionality, although a significant increase in accuracy was found by using the RFE method. In terms of classification accuracy, SVM performed the best using FS, followed by RF and KNN. Finally, only a small number of features is needed to achieve the highest performance using each classifier. This study emphasizes the benefits of rigorous FS for maximizing performance, as well as for minimizing model complexity and interpretation.

Ship Detection Based on Multiple Features in Random Forest Model for Hyperspectral Images

NASA Astrophysics Data System (ADS)

Li, N.; Ding, L.; Zhao, H.; Shi, J.; Wang, D.; Gong, X.

2018-04-01

A novel method for detecting ships which aim to make full use of both the spatial and spectral information from hyperspectral images is proposed. Firstly, the band which is high signal-noise ratio in the range of near infrared or short-wave infrared spectrum, is used to segment land and sea on Otsu threshold segmentation method. Secondly, multiple features that include spectral and texture features are extracted from hyperspectral images. Principal components analysis (PCA) is used to extract spectral features, the Grey Level Co-occurrence Matrix (GLCM) is used to extract texture features. Finally, Random Forest (RF) model is introduced to detect ships based on the extracted features. To illustrate the effectiveness of the method, we carry out experiments over the EO-1 data by comparing single feature and different multiple features. Compared with the traditional single feature method and Support Vector Machine (SVM) model, the proposed method can stably achieve the target detection of ships under complex background and can effectively improve the detection accuracy of ships.

Design of a hydraulic bending machine

Treesearch

Steven G. Hankel; Marshall Begel

2004-01-01

To keep pace with customer demands while phasing out old and unserviceable test equipment, the staff of the Engineering Mechanics Laboratory (EML) at the USDA Forest Service, Forest Products Laboratory, designed and assembled a hydraulic bending test machine. The EML built this machine to test dimension lumber, nominal 2 in. thick and up to 12 in. deep, at spans up to...

Predicting Health Care Utilization After Behavioral Health Referral Using Natural Language Processing and Machine Learning.

PubMed

Roysden, Nathaniel; Wright, Adam

2015-01-01

Mental health problems are an independent predictor of increased healthcare utilization. We created random forest classifiers for predicting two outcomes following a patient's first behavioral health encounter: decreased utilization by any amount (AUROC 0.74) and ultra-high absolute utilization (AUROC 0.88). These models may be used for clinical decision support by referring providers, to automatically detect patients who may benefit from referral, for cost management, or for risk/protection factor analysis.

A comparison of rule-based and machine learning approaches for classifying patient portal messages.

PubMed

Cronin, Robert M; Fabbri, Daniel; Denny, Joshua C; Rosenbloom, S Trent; Jackson, Gretchen Purcell

2017-09-01

Secure messaging through patient portals is an increasingly popular way that consumers interact with healthcare providers. The increasing burden of secure messaging can affect clinic staffing and workflows. Manual management of portal messages is costly and time consuming. Automated classification of portal messages could potentially expedite message triage and delivery of care. We developed automated patient portal message classifiers with rule-based and machine learning techniques using bag of words and natural language processing (NLP) approaches. To evaluate classifier performance, we used a gold standard of 3253 portal messages manually categorized using a taxonomy of communication types (i.e., main categories of informational, medical, logistical, social, and other communications, and subcategories including prescriptions, appointments, problems, tests, follow-up, contact information, and acknowledgement). We evaluated our classifiers' accuracies in identifying individual communication types within portal messages with area under the receiver-operator curve (AUC). Portal messages often contain more than one type of communication. To predict all communication types within single messages, we used the Jaccard Index. We extracted the variables of importance for the random forest classifiers. The best performing approaches to classification for the major communication types were: logistic regression for medical communications (AUC: 0.899); basic (rule-based) for informational communications (AUC: 0.842); and random forests for social communications and logistical communications (AUCs: 0.875 and 0.925, respectively). The best performing classification approach of classifiers for individual communication subtypes was random forests for Logistical-Contact Information (AUC: 0.963). The Jaccard Indices by approach were: basic classifier, Jaccard Index: 0.674; Naïve Bayes, Jaccard Index: 0.799; random forests, Jaccard Index: 0.859; and logistic regression, Jaccard Index: 0.861. For medical communications, the most predictive variables were NLP concepts (e.g., Temporal_Concept, which maps to 'morning', 'evening' and Idea_or_Concept which maps to 'appointment' and 'refill'). For logistical communications, the most predictive variables contained similar numbers of NLP variables and words (e.g., Telephone mapping to 'phone', 'insurance'). For social and informational communications, the most predictive variables were words (e.g., social: 'thanks', 'much', informational: 'question', 'mean'). This study applies automated classification methods to the content of patient portal messages and evaluates the application of NLP techniques on consumer communications in patient portal messages. We demonstrated that random forest and logistic regression approaches accurately classified the content of portal messages, although the best approach to classification varied by communication type. Words were the most predictive variables for classification of most communication types, although NLP variables were most predictive for medical communication types. As adoption of patient portals increases, automated techniques could assist in understanding and managing growing volumes of messages. Further work is needed to improve classification performance to potentially support message triage and answering. Copyright © 2017 Elsevier B.V. All rights reserved.

Estimating the impact of mineral aerosols on crop yields in food insecure regions using statistical crop models

NASA Astrophysics Data System (ADS)

Hoffman, A.; Forest, C. E.; Kemanian, A.

2016-12-01

A significant number of food-insecure nations exist in regions of the world where dust plays a large role in the climate system. While the impacts of common climate variables (e.g. temperature, precipitation, ozone, and carbon dioxide) on crop yields are relatively well understood, the impact of mineral aerosols on yields have not yet been thoroughly investigated. This research aims to develop the data and tools to progress our understanding of mineral aerosol impacts on crop yields. Suspended dust affects crop yields by altering the amount and type of radiation reaching the plant, modifying local temperature and precipitation. While dust events (i.e. dust storms) affect crop yields by depleting the soil of nutrients or by defoliation via particle abrasion. The impact of dust on yields is modeled statistically because we are uncertain which impacts will dominate the response on national and regional scales considered in this study. Multiple linear regression is used in a number of large-scale statistical crop modeling studies to estimate yield responses to various climate variables. In alignment with previous work, we develop linear crop models, but build upon this simple method of regression with machine-learning techniques (e.g. random forests) to identify important statistical predictors and isolate how dust affects yields on the scales of interest. To perform this analysis, we develop a crop-climate dataset for maize, soybean, groundnut, sorghum, rice, and wheat for the regions of West Africa, East Africa, South Africa, and the Sahel. Random forest regression models consistently model historic crop yields better than the linear models. In several instances, the random forest models accurately capture the temperature and precipitation threshold behavior in crops. Additionally, improving agricultural technology has caused a well-documented positive trend that dominates time series of global and regional yields. This trend is often removed before regression with traditional crop models, but likely at the cost of removing climate information. Our random forest models consistently discover the positive trend without removing any additional data. The application of random forests as a statistical crop model provides insight into understanding the impact of dust on yields in marginal food producing regions.

A k-mer-based barcode DNA classification methodology based on spectral representation and a neural gas network.

PubMed

Fiannaca, Antonino; La Rosa, Massimo; Rizzo, Riccardo; Urso, Alfonso

2015-07-01

In this paper, an alignment-free method for DNA barcode classification that is based on both a spectral representation and a neural gas network for unsupervised clustering is proposed. In the proposed methodology, distinctive words are identified from a spectral representation of DNA sequences. A taxonomic classification of the DNA sequence is then performed using the sequence signature, i.e., the smallest set of k-mers that can assign a DNA sequence to its proper taxonomic category. Experiments were then performed to compare our method with other supervised machine learning classification algorithms, such as support vector machine, random forest, ripper, naïve Bayes, ridor, and classification tree, which also consider short DNA sequence fragments of 200 and 300 base pairs (bp). The experimental tests were conducted over 10 real barcode datasets belonging to different animal species, which were provided by the on-line resource "Barcode of Life Database". The experimental results showed that our k-mer-based approach is directly comparable, in terms of accuracy, recall and precision metrics, with the other classifiers when considering full-length sequences. In addition, we demonstrate the robustness of our method when a classification is performed task with a set of short DNA sequences that were randomly extracted from the original data. For example, the proposed method can reach the accuracy of 64.8% at the species level with 200-bp fragments. Under the same conditions, the best other classifier (random forest) reaches the accuracy of 20.9%. Our results indicate that we obtained a clear improvement over the other classifiers for the study of short DNA barcode sequence fragments. Copyright © 2015 Elsevier B.V. All rights reserved.

Phenotyping: Using Machine Learning for Improved Pairwise Genotype Classification Based on Root Traits

PubMed Central

Zhao, Jiangsan; Bodner, Gernot; Rewald, Boris

2016-01-01

Phenotyping local crop cultivars is becoming more and more important, as they are an important genetic source for breeding – especially in regard to inherent root system architectures. Machine learning algorithms are promising tools to assist in the analysis of complex data sets; novel approaches are need to apply them on root phenotyping data of mature plants. A greenhouse experiment was conducted in large, sand-filled columns to differentiate 16 European Pisum sativum cultivars based on 36 manually derived root traits. Through combining random forest and support vector machine models, machine learning algorithms were successfully used for unbiased identification of most distinguishing root traits and subsequent pairwise cultivar differentiation. Up to 86% of pea cultivar pairs could be distinguished based on top five important root traits (Timp5) – Timp5 differed widely between cultivar pairs. Selecting top important root traits (Timp) provided a significant improved classification compared to using all available traits or randomly selected trait sets. The most frequent Timp of mature pea cultivars was total surface area of lateral roots originating from tap root segments at 0–5 cm depth. The high classification rate implies that culturing did not lead to a major loss of variability in root system architecture in the studied pea cultivars. Our results illustrate the potential of machine learning approaches for unbiased (root) trait selection and cultivar classification based on rather small, complex phenotypic data sets derived from pot experiments. Powerful statistical approaches are essential to make use of the increasing amount of (root) phenotyping information, integrating the complex trait sets describing crop cultivars. PMID:27999587

A comparison of the stochastic and machine learning approaches in hydrologic time series forecasting

NASA Astrophysics Data System (ADS)

Kim, T.; Joo, K.; Seo, J.; Heo, J. H.

2016-12-01

Hydrologic time series forecasting is an essential task in water resources management and it becomes more difficult due to the complexity of runoff process. Traditional stochastic models such as ARIMA family has been used as a standard approach in time series modeling and forecasting of hydrological variables. Due to the nonlinearity in hydrologic time series data, machine learning approaches has been studied with the advantage of discovering relevant features in a nonlinear relation among variables. This study aims to compare the predictability between the traditional stochastic model and the machine learning approach. Seasonal ARIMA model was used as the traditional time series model, and Random Forest model which consists of decision tree and ensemble method using multiple predictor approach was applied as the machine learning approach. In the application, monthly inflow data from 1986 to 2015 of Chungju dam in South Korea were used for modeling and forecasting. In order to evaluate the performances of the used models, one step ahead and multi-step ahead forecasting was applied. Root mean squared error and mean absolute error of two models were compared.

Voice based gender classification using machine learning

NASA Astrophysics Data System (ADS)

Raahul, A.; Sapthagiri, R.; Pankaj, K.; Vijayarajan, V.

2017-11-01

Gender identification is one of the major problem speech analysis today. Tracing the gender from acoustic data i.e., pitch, median, frequency etc. Machine learning gives promising results for classification problem in all the research domains. There are several performance metrics to evaluate algorithms of an area. Our Comparative model algorithm for evaluating 5 different machine learning algorithms based on eight different metrics in gender classification from acoustic data. Agenda is to identify gender, with five different algorithms: Linear Discriminant Analysis (LDA), K-Nearest Neighbour (KNN), Classification and Regression Trees (CART), Random Forest (RF), and Support Vector Machine (SVM) on basis of eight different metrics. The main parameter in evaluating any algorithms is its performance. Misclassification rate must be less in classification problems, which says that the accuracy rate must be high. Location and gender of the person have become very crucial in economic markets in the form of AdSense. Here with this comparative model algorithm, we are trying to assess the different ML algorithms and find the best fit for gender classification of acoustic data.

Random forests ensemble classifier trained with data resampling strategy to improve cardiac arrhythmia diagnosis.

PubMed

Ozçift, Akin

2011-05-01

Supervised classification algorithms are commonly used in the designing of computer-aided diagnosis systems. In this study, we present a resampling strategy based Random Forests (RF) ensemble classifier to improve diagnosis of cardiac arrhythmia. Random forests is an ensemble classifier that consists of many decision trees and outputs the class that is the mode of the class's output by individual trees. In this way, an RF ensemble classifier performs better than a single tree from classification performance point of view. In general, multiclass datasets having unbalanced distribution of sample sizes are difficult to analyze in terms of class discrimination. Cardiac arrhythmia is such a dataset that has multiple classes with small sample sizes and it is therefore adequate to test our resampling based training strategy. The dataset contains 452 samples in fourteen types of arrhythmias and eleven of these classes have sample sizes less than 15. Our diagnosis strategy consists of two parts: (i) a correlation based feature selection algorithm is used to select relevant features from cardiac arrhythmia dataset. (ii) RF machine learning algorithm is used to evaluate the performance of selected features with and without simple random sampling to evaluate the efficiency of proposed training strategy. The resultant accuracy of the classifier is found to be 90.0% and this is a quite high diagnosis performance for cardiac arrhythmia. Furthermore, three case studies, i.e., thyroid, cardiotocography and audiology, are used to benchmark the effectiveness of the proposed method. The results of experiments demonstrated the efficiency of random sampling strategy in training RF ensemble classification algorithm. Copyright © 2011 Elsevier Ltd. All rights reserved.

Accuracy of tracking forest machines with GPS

Treesearch

M.W. Veal; S.E. Taylor; T.P. McDonald; D.K. McLemore; M.R. Dunn

2001-01-01

This paper describes the results of a study that measured the accuracy of using GPS to track movement of forest machines. Two different commercially available GPS receivers (Trimble ProXR and GeoExplorer II) were used to track\\r\

Automated detection of preserved photoreceptor on optical coherence tomography in choroideremia based on machine learning.

PubMed

Wang, Zhuo; Camino, Acner; Hagag, Ahmed M; Wang, Jie; Weleber, Richard G; Yang, Paul; Pennesi, Mark E; Huang, David; Li, Dengwang; Jia, Yali

2018-05-01

Optical coherence tomography (OCT) can demonstrate early deterioration of the photoreceptor integrity caused by inherited retinal degeneration diseases (IRDs). A machine learning method based on random forests was developed to automatically detect continuous areas of preserved ellipsoid zone structure (an easily recognizable part of the photoreceptors on OCT) in 16 eyes of patients with choroideremia (a type of IRD). Pseudopodial extensions protruding from the preserved ellipsoid zone areas are detected separately by a local active contour routine. The algorithm is implemented on en face images with minimum segmentation requirements, only needing delineation of the Bruch's membrane, thus evading the inaccuracies and technical challenges associated with automatic segmentation of the ellipsoid zone in eyes with severe retinal degeneration. © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.

Osteoporosis risk prediction using machine learning and conventional methods.

PubMed

Kim, Sung Kean; Yoo, Tae Keun; Oh, Ein; Kim, Deok Won

2013-01-01

A number of clinical decision tools for osteoporosis risk assessment have been developed to select postmenopausal women for the measurement of bone mineral density. We developed and validated machine learning models with the aim of more accurately identifying the risk of osteoporosis in postmenopausal women, and compared with the ability of a conventional clinical decision tool, osteoporosis self-assessment tool (OST). We collected medical records from Korean postmenopausal women based on the Korea National Health and Nutrition Surveys (KNHANES V-1). The training data set was used to construct models based on popular machine learning algorithms such as support vector machines (SVM), random forests (RF), artificial neural networks (ANN), and logistic regression (LR) based on various predictors associated with low bone density. The learning models were compared with OST. SVM had significantly better area under the curve (AUC) of the receiver operating characteristic (ROC) than ANN, LR, and OST. Validation on the test set showed that SVM predicted osteoporosis risk with an AUC of 0.827, accuracy of 76.7%, sensitivity of 77.8%, and specificity of 76.0%. We were the first to perform comparisons of the performance of osteoporosis prediction between the machine learning and conventional methods using population-based epidemiological data. The machine learning methods may be effective tools for identifying postmenopausal women at high risk for osteoporosis.

Machine vision based quality inspection of flat glass products

NASA Astrophysics Data System (ADS)

Zauner, G.; Schagerl, M.

2014-03-01

This application paper presents a machine vision solution for the quality inspection of flat glass products. A contact image sensor (CIS) is used to generate digital images of the glass surfaces. The presented machine vision based quality inspection at the end of the production line aims to classify five different glass defect types. The defect images are usually characterized by very little `image structure', i.e. homogeneous regions without distinct image texture. Additionally, these defect images usually consist of only a few pixels. At the same time the appearance of certain defect classes can be very diverse (e.g. water drops). We used simple state-of-the-art image features like histogram-based features (std. deviation, curtosis, skewness), geometric features (form factor/elongation, eccentricity, Hu-moments) and texture features (grey level run length matrix, co-occurrence matrix) to extract defect information. The main contribution of this work now lies in the systematic evaluation of various machine learning algorithms to identify appropriate classification approaches for this specific class of images. In this way, the following machine learning algorithms were compared: decision tree (J48), random forest, JRip rules, naive Bayes, Support Vector Machine (multi class), neural network (multilayer perceptron) and k-Nearest Neighbour. We used a representative image database of 2300 defect images and applied cross validation for evaluation purposes.

Comparison of Random Forest, k-Nearest Neighbor, and Support Vector Machine Classifiers for Land Cover Classification Using Sentinel-2 Imagery

PubMed Central

Thanh Noi, Phan; Kappas, Martin

2017-01-01

In previous classification studies, three non-parametric classifiers, Random Forest (RF), k-Nearest Neighbor (kNN), and Support Vector Machine (SVM), were reported as the foremost classifiers at producing high accuracies. However, only a few studies have compared the performances of these classifiers with different training sample sizes for the same remote sensing images, particularly the Sentinel-2 Multispectral Imager (MSI). In this study, we examined and compared the performances of the RF, kNN, and SVM classifiers for land use/cover classification using Sentinel-2 image data. An area of 30 × 30 km2 within the Red River Delta of Vietnam with six land use/cover types was classified using 14 different training sample sizes, including balanced and imbalanced, from 50 to over 1250 pixels/class. All classification results showed a high overall accuracy (OA) ranging from 90% to 95%. Among the three classifiers and 14 sub-datasets, SVM produced the highest OA with the least sensitivity to the training sample sizes, followed consecutively by RF and kNN. In relation to the sample size, all three classifiers showed a similar and high OA (over 93.85%) when the training sample size was large enough, i.e., greater than 750 pixels/class or representing an area of approximately 0.25% of the total study area. The high accuracy was achieved with both imbalanced and balanced datasets. PMID:29271909

Comparison of Random Forest, k-Nearest Neighbor, and Support Vector Machine Classifiers for Land Cover Classification Using Sentinel-2 Imagery.

PubMed

Thanh Noi, Phan; Kappas, Martin

2017-12-22

In previous classification studies, three non-parametric classifiers, Random Forest (RF), k-Nearest Neighbor (kNN), and Support Vector Machine (SVM), were reported as the foremost classifiers at producing high accuracies. However, only a few studies have compared the performances of these classifiers with different training sample sizes for the same remote sensing images, particularly the Sentinel-2 Multispectral Imager (MSI). In this study, we examined and compared the performances of the RF, kNN, and SVM classifiers for land use/cover classification using Sentinel-2 image data. An area of 30 × 30 km² within the Red River Delta of Vietnam with six land use/cover types was classified using 14 different training sample sizes, including balanced and imbalanced, from 50 to over 1250 pixels/class. All classification results showed a high overall accuracy (OA) ranging from 90% to 95%. Among the three classifiers and 14 sub-datasets, SVM produced the highest OA with the least sensitivity to the training sample sizes, followed consecutively by RF and kNN. In relation to the sample size, all three classifiers showed a similar and high OA (over 93.85%) when the training sample size was large enough, i.e., greater than 750 pixels/class or representing an area of approximately 0.25% of the total study area. The high accuracy was achieved with both imbalanced and balanced datasets.

Discrimination of raw and processed Dipsacus asperoides by near infrared spectroscopy combined with least squares-support vector machine and random forests

NASA Astrophysics Data System (ADS)

Xin, Ni; Gu, Xiao-Feng; Wu, Hao; Hu, Yu-Zhu; Yang, Zhong-Lin

2012-04-01

Most herbal medicines could be processed to fulfill the different requirements of therapy. The purpose of this study was to discriminate between raw and processed Dipsacus asperoides, a common traditional Chinese medicine, based on their near infrared (NIR) spectra. Least squares-support vector machine (LS-SVM) and random forests (RF) were employed for full-spectrum classification. Three types of kernels, including linear kernel, polynomial kernel and radial basis function kernel (RBF), were checked for optimization of LS-SVM model. For comparison, a linear discriminant analysis (LDA) model was performed for classification, and the successive projections algorithm (SPA) was executed prior to building an LDA model to choose an appropriate subset of wavelengths. The three methods were applied to a dataset containing 40 raw herbs and 40 corresponding processed herbs. We ran 50 runs of 10-fold cross validation to evaluate the model's efficiency. The performance of the LS-SVM with RBF kernel (RBF LS-SVM) was better than the other two kernels. The RF, RBF LS-SVM and SPA-LDA successfully classified all test samples. The mean error rates for the 50 runs of 10-fold cross validation were 1.35% for RBF LS-SVM, 2.87% for RF, and 2.50% for SPA-LDA. The best classification results were obtained by using LS-SVM with RBF kernel, while RF was fast in the training and making predictions.

Comparative Performance Analysis of Support Vector Machine, Random Forest, Logistic Regression and k-Nearest Neighbours in Rainbow Trout (Oncorhynchus Mykiss) Classification Using Image-Based Features

PubMed Central

Císař, Petr; Labbé, Laurent; Souček, Pavel; Pelissier, Pablo; Kerneis, Thierry

2018-01-01

The main aim of this study was to develop a new objective method for evaluating the impacts of different diets on the live fish skin using image-based features. In total, one-hundred and sixty rainbow trout (Oncorhynchus mykiss) were fed either a fish-meal based diet (80 fish) or a 100% plant-based diet (80 fish) and photographed using consumer-grade digital camera. Twenty-three colour features and four texture features were extracted. Four different classification methods were used to evaluate fish diets including Random forest (RF), Support vector machine (SVM), Logistic regression (LR) and k-Nearest neighbours (k-NN). The SVM with radial based kernel provided the best classifier with correct classification rate (CCR) of 82% and Kappa coefficient of 0.65. Although the both LR and RF methods were less accurate than SVM, they achieved good classification with CCR 75% and 70% respectively. The k-NN was the least accurate (40%) classification model. Overall, it can be concluded that consumer-grade digital cameras could be employed as the fast, accurate and non-invasive sensor for classifying rainbow trout based on their diets. Furthermore, these was a close association between image-based features and fish diet received during cultivation. These procedures can be used as non-invasive, accurate and precise approaches for monitoring fish status during the cultivation by evaluating diet’s effects on fish skin. PMID:29596375

Comparative Performance Analysis of Support Vector Machine, Random Forest, Logistic Regression and k-Nearest Neighbours in Rainbow Trout (Oncorhynchus Mykiss) Classification Using Image-Based Features.

PubMed

Saberioon, Mohammadmehdi; Císař, Petr; Labbé, Laurent; Souček, Pavel; Pelissier, Pablo; Kerneis, Thierry

2018-03-29

The main aim of this study was to develop a new objective method for evaluating the impacts of different diets on the live fish skin using image-based features. In total, one-hundred and sixty rainbow trout ( Oncorhynchus mykiss ) were fed either a fish-meal based diet (80 fish) or a 100% plant-based diet (80 fish) and photographed using consumer-grade digital camera. Twenty-three colour features and four texture features were extracted. Four different classification methods were used to evaluate fish diets including Random forest (RF), Support vector machine (SVM), Logistic regression (LR) and k -Nearest neighbours ( k -NN). The SVM with radial based kernel provided the best classifier with correct classification rate (CCR) of 82% and Kappa coefficient of 0.65. Although the both LR and RF methods were less accurate than SVM, they achieved good classification with CCR 75% and 70% respectively. The k -NN was the least accurate (40%) classification model. Overall, it can be concluded that consumer-grade digital cameras could be employed as the fast, accurate and non-invasive sensor for classifying rainbow trout based on their diets. Furthermore, these was a close association between image-based features and fish diet received during cultivation. These procedures can be used as non-invasive, accurate and precise approaches for monitoring fish status during the cultivation by evaluating diet's effects on fish skin.

GIS-based groundwater potential mapping using boosted regression tree, classification and regression tree, and random forest machine learning models in Iran.

PubMed

Naghibi, Seyed Amir; Pourghasemi, Hamid Reza; Dixon, Barnali

2016-01-01

Groundwater is considered one of the most valuable fresh water resources. The main objective of this study was to produce groundwater spring potential maps in the Koohrang Watershed, Chaharmahal-e-Bakhtiari Province, Iran, using three machine learning models: boosted regression tree (BRT), classification and regression tree (CART), and random forest (RF). Thirteen hydrological-geological-physiographical (HGP) factors that influence locations of springs were considered in this research. These factors include slope degree, slope aspect, altitude, topographic wetness index (TWI), slope length (LS), plan curvature, profile curvature, distance to rivers, distance to faults, lithology, land use, drainage density, and fault density. Subsequently, groundwater spring potential was modeled and mapped using CART, RF, and BRT algorithms. The predicted results from the three models were validated using the receiver operating characteristics curve (ROC). From 864 springs identified, 605 (≈70 %) locations were used for the spring potential mapping, while the remaining 259 (≈30 %) springs were used for the model validation. The area under the curve (AUC) for the BRT model was calculated as 0.8103 and for CART and RF the AUC were 0.7870 and 0.7119, respectively. Therefore, it was concluded that the BRT model produced the best prediction results while predicting locations of springs followed by CART and RF models, respectively. Geospatially integrated BRT, CART, and RF methods proved to be useful in generating the spring potential map (SPM) with reasonable accuracy.

Radiomic modeling of BI-RADS density categories

NASA Astrophysics Data System (ADS)

Wei, Jun; Chan, Heang-Ping; Helvie, Mark A.; Roubidoux, Marilyn A.; Zhou, Chuan; Hadjiiski, Lubomir

2017-03-01

Screening mammography is the most effective and low-cost method to date for early cancer detection. Mammographic breast density has been shown to be highly correlated with breast cancer risk. We are developing a radiomic model for BI-RADS density categorization on digital mammography (FFDM) with a supervised machine learning approach. With IRB approval, we retrospectively collected 478 FFDMs from 478 women. As a gold standard, breast density was assessed by an MQSA radiologist based on BI-RADS categories. The raw FFDMs were used for computerized density assessment. The raw FFDM first underwent log-transform to approximate the x-ray sensitometric response, followed by multiscale processing to enhance the fibroglandular densities and parenchymal patterns. Three ROIs were automatically identified based on the keypoint distribution, where the keypoints were obtained as the extrema in the image Gaussian scale-space. A total of 73 features, including intensity and texture features that describe the density and the parenchymal pattern, were extracted from each breast. Our BI-RADS density estimator was constructed by using a random forest classifier. We used a 10-fold cross validation resampling approach to estimate the errors. With the random forest classifier, computerized density categories for 412 of the 478 cases agree with radiologist's assessment (weighted kappa = 0.93). The machine learning method with radiomic features as predictors demonstrated a high accuracy in classifying FFDMs into BI-RADS density categories. Further work is underway to improve our system performance as well as to perform an independent testing using a large unseen FFDM set.

23. Photo copy of photograph, (original in Forest Service Office, ...

Library of Congress Historic Buildings Survey, Historic Engineering Record, Historic Landscapes Survey

23. Photo copy of photograph, (original in Forest Service Office, Elkins, WV, photo #248336, 'Tree nurseries-seed bed seeding machine'), D. A. Oliver, 1930. VIEW WEST, SEEDING MACHINE. - Parsons Nursery, South side of U.S. Route 219, Parsons, Tucker County, WV

Risk estimation using probability machines.

PubMed

Dasgupta, Abhijit; Szymczak, Silke; Moore, Jason H; Bailey-Wilson, Joan E; Malley, James D

2014-03-01

Logistic regression has been the de facto, and often the only, model used in the description and analysis of relationships between a binary outcome and observed features. It is widely used to obtain the conditional probabilities of the outcome given predictors, as well as predictor effect size estimates using conditional odds ratios. We show how statistical learning machines for binary outcomes, provably consistent for the nonparametric regression problem, can be used to provide both consistent conditional probability estimation and conditional effect size estimates. Effect size estimates from learning machines leverage our understanding of counterfactual arguments central to the interpretation of such estimates. We show that, if the data generating model is logistic, we can recover accurate probability predictions and effect size estimates with nearly the same efficiency as a correct logistic model, both for main effects and interactions. We also propose a method using learning machines to scan for possible interaction effects quickly and efficiently. Simulations using random forest probability machines are presented. The models we propose make no assumptions about the data structure, and capture the patterns in the data by just specifying the predictors involved and not any particular model structure. So they do not run the same risks of model mis-specification and the resultant estimation biases as a logistic model. This methodology, which we call a "risk machine", will share properties from the statistical machine that it is derived from.

Machine Learning for Detecting Gene-Gene Interactions

PubMed Central

McKinney, Brett A.; Reif, David M.; Ritchie, Marylyn D.; Moore, Jason H.

2011-01-01

Complex interactions among genes and environmental factors are known to play a role in common human disease aetiology. There is a growing body of evidence to suggest that complex interactions are ‘the norm’ and, rather than amounting to a small perturbation to classical Mendelian genetics, interactions may be the predominant effect. Traditional statistical methods are not well suited for detecting such interactions, especially when the data are high dimensional (many attributes or independent variables) or when interactions occur between more than two polymorphisms. In this review, we discuss machine-learning models and algorithms for identifying and characterising susceptibility genes in common, complex, multifactorial human diseases. We focus on the following machine-learning methods that have been used to detect gene-gene interactions: neural networks, cellular automata, random forests, and multifactor dimensionality reduction. We conclude with some ideas about how these methods and others can be integrated into a comprehensive and flexible framework for data mining and knowledge discovery in human genetics. PMID:16722772

Modeling Music Emotion Judgments Using Machine Learning Methods

PubMed Central

Vempala, Naresh N.; Russo, Frank A.

2018-01-01

Emotion judgments and five channels of physiological data were obtained from 60 participants listening to 60 music excerpts. Various machine learning (ML) methods were used to model the emotion judgments inclusive of neural networks, linear regression, and random forests. Input for models of perceived emotion consisted of audio features extracted from the music recordings. Input for models of felt emotion consisted of physiological features extracted from the physiological recordings. Models were trained and interpreted with consideration of the classic debate in music emotion between cognitivists and emotivists. Our models supported a hybrid position wherein emotion judgments were influenced by a combination of perceived and felt emotions. In comparing the different ML approaches that were used for modeling, we conclude that neural networks were optimal, yielding models that were flexible as well as interpretable. Inspection of a committee machine, encompassing an ensemble of networks, revealed that arousal judgments were predominantly influenced by felt emotion, whereas valence judgments were predominantly influenced by perceived emotion. PMID:29354080

Framework for Infectious Disease Analysis: A comprehensive and integrative multi-modeling approach to disease prediction and management.

PubMed

Erraguntla, Madhav; Zapletal, Josef; Lawley, Mark

2017-12-01

The impact of infectious disease on human populations is a function of many factors including environmental conditions, vector dynamics, transmission mechanics, social and cultural behaviors, and public policy. A comprehensive framework for disease management must fully connect the complete disease lifecycle, including emergence from reservoir populations, zoonotic vector transmission, and impact on human societies. The Framework for Infectious Disease Analysis is a software environment and conceptual architecture for data integration, situational awareness, visualization, prediction, and intervention assessment. Framework for Infectious Disease Analysis automatically collects biosurveillance data using natural language processing, integrates structured and unstructured data from multiple sources, applies advanced machine learning, and uses multi-modeling for analyzing disease dynamics and testing interventions in complex, heterogeneous populations. In the illustrative case studies, natural language processing from social media, news feeds, and websites was used for information extraction, biosurveillance, and situation awareness. Classification machine learning algorithms (support vector machines, random forests, and boosting) were used for disease predictions.

Modeling Music Emotion Judgments Using Machine Learning Methods.

PubMed

Vempala, Naresh N; Russo, Frank A

2017-01-01

Emotion judgments and five channels of physiological data were obtained from 60 participants listening to 60 music excerpts. Various machine learning (ML) methods were used to model the emotion judgments inclusive of neural networks, linear regression, and random forests. Input for models of perceived emotion consisted of audio features extracted from the music recordings. Input for models of felt emotion consisted of physiological features extracted from the physiological recordings. Models were trained and interpreted with consideration of the classic debate in music emotion between cognitivists and emotivists. Our models supported a hybrid position wherein emotion judgments were influenced by a combination of perceived and felt emotions. In comparing the different ML approaches that were used for modeling, we conclude that neural networks were optimal, yielding models that were flexible as well as interpretable. Inspection of a committee machine, encompassing an ensemble of networks, revealed that arousal judgments were predominantly influenced by felt emotion, whereas valence judgments were predominantly influenced by perceived emotion.

SU-D-204-06: Integration of Machine Learning and Bioinformatics Methods to Analyze Genome-Wide Association Study Data for Rectal Bleeding and Erectile Dysfunction Following Radiotherapy in Prostate Cancer

DOE Office of Scientific and Technical Information (OSTI.GOV)

Oh, J; Deasy, J; Kerns, S

Purpose: We investigated whether integration of machine learning and bioinformatics techniques on genome-wide association study (GWAS) data can improve the performance of predictive models in predicting the risk of developing radiation-induced late rectal bleeding and erectile dysfunction in prostate cancer patients. Methods: We analyzed a GWAS dataset generated from 385 prostate cancer patients treated with radiotherapy. Using genotype information from these patients, we designed a machine learning-based predictive model of late radiation-induced toxicities: rectal bleeding and erectile dysfunction. The model building process was performed using 2/3 of samples (training) and the predictive model was tested with 1/3 of samples (validation).more » To identify important single nucleotide polymorphisms (SNPs), we computed the SNP importance score, resulting from our random forest regression model. We performed gene ontology (GO) enrichment analysis for nearby genes of the important SNPs. Results: After univariate analysis on the training dataset, we filtered out many SNPs with p>0.001, resulting in 749 and 367 SNPs that were used in the model building process for rectal bleeding and erectile dysfunction, respectively. On the validation dataset, our random forest regression model achieved the area under the curve (AUC)=0.70 and 0.62 for rectal bleeding and erectile dysfunction, respectively. We performed GO enrichment analysis for the top 25%, 50%, 75%, and 100% SNPs out of the select SNPs in the univariate analysis. When we used the top 50% SNPs, more plausible biological processes were obtained for both toxicities. An additional test with the top 50% SNPs improved predictive power with AUC=0.71 and 0.65 for rectal bleeding and erectile dysfunction. A better performance was achieved with AUC=0.67 when age and androgen deprivation therapy were added to the model for erectile dysfunction. Conclusion: Our approach that combines machine learning and bioinformatics techniques enabled designing better models and identifying more plausible biological processes associated with the outcomes.« less

Benchmark of Machine Learning Methods for Classification of a SENTINEL-2 Image

NASA Astrophysics Data System (ADS)

Pirotti, F.; Sunar, F.; Piragnolo, M.

2016-06-01

Thanks to mainly ESA and USGS, a large bulk of free images of the Earth is readily available nowadays. One of the main goals of remote sensing is to label images according to a set of semantic categories, i.e. image classification. This is a very challenging issue since land cover of a specific class may present a large spatial and spectral variability and objects may appear at different scales and orientations. In this study, we report the results of benchmarking 9 machine learning algorithms tested for accuracy and speed in training and classification of land-cover classes in a Sentinel-2 dataset. The following machine learning methods (MLM) have been tested: linear discriminant analysis, k-nearest neighbour, random forests, support vector machines, multi layered perceptron, multi layered perceptron ensemble, ctree, boosting, logarithmic regression. The validation is carried out using a control dataset which consists of an independent classification in 11 land-cover classes of an area about 60 km2, obtained by manual visual interpretation of high resolution images (20 cm ground sampling distance) by experts. In this study five out of the eleven classes are used since the others have too few samples (pixels) for testing and validating subsets. The classes used are the following: (i) urban (ii) sowable areas (iii) water (iv) tree plantations (v) grasslands. Validation is carried out using three different approaches: (i) using pixels from the training dataset (train), (ii) using pixels from the training dataset and applying cross-validation with the k-fold method (kfold) and (iii) using all pixels from the control dataset. Five accuracy indices are calculated for the comparison between the values predicted with each model and control values over three sets of data: the training dataset (train), the whole control dataset (full) and with k-fold cross-validation (kfold) with ten folds. Results from validation of predictions of the whole dataset (full) show the random forests method with the highest values; kappa index ranging from 0.55 to 0.42 respectively with the most and least number pixels for training. The two neural networks (multi layered perceptron and its ensemble) and the support vector machines - with default radial basis function kernel - methods follow closely with comparable performance.

BitterSweetForest: A random forest based binary classifier to predict bitterness and sweetness of chemical compounds

NASA Astrophysics Data System (ADS)

Banerjee, Priyanka; Preissner, Robert

2018-04-01

Taste of a chemical compounds present in food stimulates us to take in nutrients and avoid poisons. However, the perception of taste greatly depends on the genetic as well as evolutionary perspectives. The aim of this work was the development and validation of a machine learning model based on molecular fingerprints to discriminate between sweet and bitter taste of molecules. BitterSweetForest is the first open access model based on KNIME workflow that provides platform for prediction of bitter and sweet taste of chemical compounds using molecular fingerprints and Random Forest based classifier. The constructed model yielded an accuracy of 95% and an AUC of 0.98 in cross-validation. In independent test set, BitterSweetForest achieved an accuracy of 96 % and an AUC of 0.98 for bitter and sweet taste prediction. The constructed model was further applied to predict the bitter and sweet taste of natural compounds, approved drugs as well as on an acute toxicity compound data set. BitterSweetForest suggests 70% of the natural product space, as bitter and 10 % of the natural product space as sweet with confidence score of 0.60 and above. 77 % of the approved drug set was predicted as bitter and 2% as sweet with a confidence scores of 0.75 and above. Similarly, 75% of the total compounds from acute oral toxicity class were predicted only as bitter with a minimum confidence score of 0.75, revealing toxic compounds are mostly bitter. Furthermore, we applied a Bayesian based feature analysis method to discriminate the most occurring chemical features between sweet and bitter compounds from the feature space of a circular fingerprint.

BitterSweetForest: A Random Forest Based Binary Classifier to Predict Bitterness and Sweetness of Chemical Compounds

PubMed Central

Banerjee, Priyanka; Preissner, Robert

2018-01-01

Taste of a chemical compound present in food stimulates us to take in nutrients and avoid poisons. However, the perception of taste greatly depends on the genetic as well as evolutionary perspectives. The aim of this work was the development and validation of a machine learning model based on molecular fingerprints to discriminate between sweet and bitter taste of molecules. BitterSweetForest is the first open access model based on KNIME workflow that provides platform for prediction of bitter and sweet taste of chemical compounds using molecular fingerprints and Random Forest based classifier. The constructed model yielded an accuracy of 95% and an AUC of 0.98 in cross-validation. In independent test set, BitterSweetForest achieved an accuracy of 96% and an AUC of 0.98 for bitter and sweet taste prediction. The constructed model was further applied to predict the bitter and sweet taste of natural compounds, approved drugs as well as on an acute toxicity compound data set. BitterSweetForest suggests 70% of the natural product space, as bitter and 10% of the natural product space as sweet with confidence score of 0.60 and above. 77% of the approved drug set was predicted as bitter and 2% as sweet with a confidence score of 0.75 and above. Similarly, 75% of the total compounds from acute oral toxicity class were predicted only as bitter with a minimum confidence score of 0.75, revealing toxic compounds are mostly bitter. Furthermore, we applied a Bayesian based feature analysis method to discriminate the most occurring chemical features between sweet and bitter compounds using the feature space of a circular fingerprint. PMID:29696137

Exploring diversity in ensemble classification: Applications in large area land cover mapping

NASA Astrophysics Data System (ADS)

Mellor, Andrew; Boukir, Samia

2017-07-01

Ensemble classifiers, such as random forests, are now commonly applied in the field of remote sensing, and have been shown to perform better than single classifier systems, resulting in reduced generalisation error. Diversity across the members of ensemble classifiers is known to have a strong influence on classification performance - whereby classifier errors are uncorrelated and more uniformly distributed across ensemble members. The relationship between ensemble diversity and classification performance has not yet been fully explored in the fields of information science and machine learning and has never been examined in the field of remote sensing. This study is a novel exploration of ensemble diversity and its link to classification performance, applied to a multi-class canopy cover classification problem using random forests and multisource remote sensing and ancillary GIS data, across seven million hectares of diverse dry-sclerophyll dominated public forests in Victoria Australia. A particular emphasis is placed on analysing the relationship between ensemble diversity and ensemble margin - two key concepts in ensemble learning. The main novelty of our work is on boosting diversity by emphasizing the contribution of lower margin instances used in the learning process. Exploring the influence of tree pruning on diversity is also a new empirical analysis that contributes to a better understanding of ensemble performance. Results reveal insights into the trade-off between ensemble classification accuracy and diversity, and through the ensemble margin, demonstrate how inducing diversity by targeting lower margin training samples is a means of achieving better classifier performance for more difficult or rarer classes and reducing information redundancy in classification problems. Our findings inform strategies for collecting training data and designing and parameterising ensemble classifiers, such as random forests. This is particularly important in large area remote sensing applications, for which training data is costly and resource intensive to collect.

Why choose Random Forest to predict rare species distribution with few samples in large undersampled areas? Three Asian crane species models provide supporting evidence.

PubMed

Mi, Chunrong; Huettmann, Falk; Guo, Yumin; Han, Xuesong; Wen, Lijia

2017-01-01

Species distribution models (SDMs) have become an essential tool in ecology, biogeography, evolution and, more recently, in conservation biology. How to generalize species distributions in large undersampled areas, especially with few samples, is a fundamental issue of SDMs. In order to explore this issue, we used the best available presence records for the Hooded Crane ( Grus monacha , n  = 33), White-naped Crane ( Grus vipio , n  = 40), and Black-necked Crane ( Grus nigricollis , n  = 75) in China as three case studies, employing four powerful and commonly used machine learning algorithms to map the breeding distributions of the three species: TreeNet (Stochastic Gradient Boosting, Boosted Regression Tree Model), Random Forest, CART (Classification and Regression Tree) and Maxent (Maximum Entropy Models). In addition, we developed an ensemble forecast by averaging predicted probability of the above four models results. Commonly used model performance metrics (Area under ROC (AUC) and true skill statistic (TSS)) were employed to evaluate model accuracy. The latest satellite tracking data and compiled literature data were used as two independent testing datasets to confront model predictions. We found Random Forest demonstrated the best performance for the most assessment method, provided a better model fit to the testing data, and achieved better species range maps for each crane species in undersampled areas. Random Forest has been generally available for more than 20 years and has been known to perform extremely well in ecological predictions. However, while increasingly on the rise, its potential is still widely underused in conservation, (spatial) ecological applications and for inference. Our results show that it informs ecological and biogeographical theories as well as being suitable for conservation applications, specifically when the study area is undersampled. This method helps to save model-selection time and effort, and allows robust and rapid assessments and decisions for efficient conservation.

Why choose Random Forest to predict rare species distribution with few samples in large undersampled areas? Three Asian crane species models provide supporting evidence

PubMed Central

Mi, Chunrong; Huettmann, Falk; Han, Xuesong; Wen, Lijia

2017-01-01

Species distribution models (SDMs) have become an essential tool in ecology, biogeography, evolution and, more recently, in conservation biology. How to generalize species distributions in large undersampled areas, especially with few samples, is a fundamental issue of SDMs. In order to explore this issue, we used the best available presence records for the Hooded Crane (Grus monacha, n = 33), White-naped Crane (Grus vipio, n = 40), and Black-necked Crane (Grus nigricollis, n = 75) in China as three case studies, employing four powerful and commonly used machine learning algorithms to map the breeding distributions of the three species: TreeNet (Stochastic Gradient Boosting, Boosted Regression Tree Model), Random Forest, CART (Classification and Regression Tree) and Maxent (Maximum Entropy Models). In addition, we developed an ensemble forecast by averaging predicted probability of the above four models results. Commonly used model performance metrics (Area under ROC (AUC) and true skill statistic (TSS)) were employed to evaluate model accuracy. The latest satellite tracking data and compiled literature data were used as two independent testing datasets to confront model predictions. We found Random Forest demonstrated the best performance for the most assessment method, provided a better model fit to the testing data, and achieved better species range maps for each crane species in undersampled areas. Random Forest has been generally available for more than 20 years and has been known to perform extremely well in ecological predictions. However, while increasingly on the rise, its potential is still widely underused in conservation, (spatial) ecological applications and for inference. Our results show that it informs ecological and biogeographical theories as well as being suitable for conservation applications, specifically when the study area is undersampled. This method helps to save model-selection time and effort, and allows robust and rapid assessments and decisions for efficient conservation. PMID:28097060

An application of quantile random forests for predictive mapping of forest attributes

Treesearch

E.A. Freeman; G.G. Moisen

2015-01-01

Increasingly, random forest models are used in predictive mapping of forest attributes. Traditional random forests output the mean prediction from the random trees. Quantile regression forests (QRF) is an extension of random forests developed by Nicolai Meinshausen that provides non-parametric estimates of the median predicted value as well as prediction quantiles. It...

Reducing RANS Model Error Using Random Forest

NASA Astrophysics Data System (ADS)

Wang, Jian-Xun; Wu, Jin-Long; Xiao, Heng; Ling, Julia

2016-11-01

Reynolds-Averaged Navier-Stokes (RANS) models are still the work-horse tools in the turbulence modeling of industrial flows. However, the model discrepancy due to the inadequacy of modeled Reynolds stresses largely diminishes the reliability of simulation results. In this work we use a physics-informed machine learning approach to improve the RANS modeled Reynolds stresses and propagate them to obtain the mean velocity field. Specifically, the functional forms of Reynolds stress discrepancies with respect to mean flow features are trained based on an offline database of flows with similar characteristics. The random forest model is used to predict Reynolds stress discrepancies in new flows. Then the improved Reynolds stresses are propagated to the velocity field via RANS equations. The effects of expanding the feature space through the use of a complete basis of Galilean tensor invariants are also studied. The flow in a square duct, which is challenging for standard RANS models, is investigated to demonstrate the merit of the proposed approach. The results show that both the Reynolds stresses and the propagated velocity field are improved over the baseline RANS predictions. SAND Number: SAND2016-7437 A

Random forest classification of stars in the Galactic Centre

NASA Astrophysics Data System (ADS)

Plewa, P. M.

2018-05-01

Near-infrared high-angular resolution imaging observations of the Milky Way's nuclear star cluster have revealed all luminous members of the existing stellar population within the central parsec. Generally, these stars are either evolved late-type giants or massive young, early-type stars. We revisit the problem of stellar classification based on intermediate-band photometry in the K band, with the primary aim of identifying faint early-type candidate stars in the extended vicinity of the central massive black hole. A random forest classifier, trained on a subsample of spectroscopically identified stars, performs similarly well as competitive methods (F1 = 0.85), without involving any model of stellar spectral energy distributions. Advantages of using such a machine-trained classifier are a minimum of required calibration effort, a predictive accuracy expected to improve as more training data become available, and the ease of application to future, larger data sets. By applying this classifier to archive data, we are also able to reproduce the results of previous studies of the spatial distribution and the K-band luminosity function of both the early- and late-type stars.

c-Fos expression predicts long-term social memory retrieval in mice.

PubMed

Lüscher Dias, Thomaz; Fernandes Golino, Hudson; Moura de Oliveira, Vinícius Elias; Dutra Moraes, Márcio Flávio; Schenatto Pereira, Grace

2016-10-15

The way the rodent brain generally processes socially relevant information is rather well understood. How social information is stored into long-term social memory, however, is still under debate. Here, brain c-Fos expression was measured after adult mice were exposed to familiar or novel juveniles and expression was compared in several memory and socially relevant brain areas. Machine Learning algorithm Random Forest was then used to predict the social interaction category of adult mice based on c-Fos expression in these areas. Interaction with a familiar co-specific altered brain activation in the olfactory bulb, amygdala, hippocampus, lateral septum and medial prefrontal cortex. Remarkably, Random Forest was able to predict interaction with a familiar juvenile with 100% accuracy. Activity in the olfactory bulb, amygdala, hippocampus and the medial prefrontal cortex were crucial to this prediction. From our results, we suggest long-term social memory depends on initial social olfactory processing in the medial amygdala and its output connections synergistically with non-social contextual integration by the hippocampus and medial prefrontal cortex top-down modulation of primary olfactory structures. Copyright © 2016 Elsevier B.V. All rights reserved.

Quantifying and Characterizing Tonic Thermal Pain Across Subjects From EEG Data Using Random Forest Models.

PubMed

Vijayakumar, Vishal; Case, Michelle; Shirinpour, Sina; He, Bin

2017-12-01

Effective pain assessment and management strategies are needed to better manage pain. In addition to self-report, an objective pain assessment system can provide a more complete picture of the neurophysiological basis for pain. In this study, a robust and accurate machine learning approach is developed to quantify tonic thermal pain across healthy subjects into a maximum of ten distinct classes. A random forest model was trained to predict pain scores using time-frequency wavelet representations of independent components obtained from electroencephalography (EEG) data, and the relative importance of each frequency band to pain quantification is assessed. The mean classification accuracy for predicting pain on an independent test subject for a range of 1-10 is 89.45%, highest among existing state of the art quantification algorithms for EEG. The gamma band is the most important to both intersubject and intrasubject classification accuracy. The robustness and generalizability of the classifier are demonstrated. Our results demonstrate the potential of this tool to be used clinically to help us to improve chronic pain treatment and establish spectral biomarkers for future pain-related studies using EEG.

Indonesian name matching using machine learning supervised approach

NASA Astrophysics Data System (ADS)

Alifikri, Mohamad; Arif Bijaksana, Moch.

2018-03-01

Most existing name matching methods are developed for English language and so they cover the characteristics of this language. Up to this moment, there is no specific one has been designed and implemented for Indonesian names. The purpose of this thesis is to develop Indonesian name matching dataset as a contribution to academic research and to propose suitable feature set by utilizing combination of context of name strings and its permute-winkler score. Machine learning classification algorithms is taken as the method for performing name matching. Based on the experiments, by using tuned Random Forest algorithm and proposed features, there is an improvement of matching performance by approximately 1.7% and it is able to reduce until 70% misclassification result of the state of the arts methods. This improving performance makes the matching system more effective and reduces the risk of misclassified matches.

Development and implementation of (Q)SAR modeling within the CHARMMing web-user interface.

PubMed

Weidlich, Iwona E; Pevzner, Yuri; Miller, Benjamin T; Filippov, Igor V; Woodcock, H Lee; Brooks, Bernard R

2015-01-05

Recent availability of large publicly accessible databases of chemical compounds and their biological activities (PubChem, ChEMBL) has inspired us to develop a web-based tool for structure activity relationship and quantitative structure activity relationship modeling to add to the services provided by CHARMMing (www.charmming.org). This new module implements some of the most recent advances in modern machine learning algorithms-Random Forest, Support Vector Machine, Stochastic Gradient Descent, Gradient Tree Boosting, so forth. A user can import training data from Pubchem Bioassay data collections directly from our interface or upload his or her own SD files which contain structures and activity information to create new models (either categorical or numerical). A user can then track the model generation process and run models on new data to predict activity. © 2014 Wiley Periodicals, Inc.

Beyond where to how: a machine learning approach for sensing mobility contexts using smartphone sensors.

PubMed

Guinness, Robert E

2015-04-28

This paper presents the results of research on the use of smartphone sensors (namely, GPS and accelerometers), geospatial information (points of interest, such as bus stops and train stations) and machine learning (ML) to sense mobility contexts. Our goal is to develop techniques to continuously and automatically detect a smartphone user's mobility activities, including walking, running, driving and using a bus or train, in real-time or near-real-time (<5 s). We investigated a wide range of supervised learning techniques for classification, including decision trees (DT), support vector machines (SVM), naive Bayes classifiers (NB), Bayesian networks (BN), logistic regression (LR), artificial neural networks (ANN) and several instance-based classifiers (KStar, LWLand IBk). Applying ten-fold cross-validation, the best performers in terms of correct classification rate (i.e., recall) were DT (96.5%), BN (90.9%), LWL (95.5%) and KStar (95.6%). In particular, the DT-algorithm RandomForest exhibited the best overall performance. After a feature selection process for a subset of algorithms, the performance was improved slightly. Furthermore, after tuning the parameters of RandomForest, performance improved to above 97.5%. Lastly, we measured the computational complexity of the classifiers, in terms of central processing unit (CPU) time needed for classification, to provide a rough comparison between the algorithms in terms of battery usage requirements. As a result, the classifiers can be ranked from lowest to highest complexity (i.e., computational cost) as follows: SVM, ANN, LR, BN, DT, NB, IBk, LWL and KStar. The instance-based classifiers take considerably more computational time than the non-instance-based classifiers, whereas the slowest non-instance-based classifier (NB) required about five-times the amount of CPU time as the fastest classifier (SVM). The above results suggest that DT algorithms are excellent candidates for detecting mobility contexts in smartphones, both in terms of performance and computational complexity.

Beyond Where to How: A Machine Learning Approach for Sensing Mobility Contexts Using Smartphone Sensors †

PubMed Central

Guinness, Robert E.

2015-01-01

This paper presents the results of research on the use of smartphone sensors (namely, GPS and accelerometers), geospatial information (points of interest, such as bus stops and train stations) and machine learning (ML) to sense mobility contexts. Our goal is to develop techniques to continuously and automatically detect a smartphone user's mobility activities, including walking, running, driving and using a bus or train, in real-time or near-real-time (<5 s). We investigated a wide range of supervised learning techniques for classification, including decision trees (DT), support vector machines (SVM), naive Bayes classifiers (NB), Bayesian networks (BN), logistic regression (LR), artificial neural networks (ANN) and several instance-based classifiers (KStar, LWLand IBk). Applying ten-fold cross-validation, the best performers in terms of correct classification rate (i.e., recall) were DT (96.5%), BN (90.9%), LWL (95.5%) and KStar (95.6%). In particular, the DT-algorithm RandomForest exhibited the best overall performance. After a feature selection process for a subset of algorithms, the performance was improved slightly. Furthermore, after tuning the parameters of RandomForest, performance improved to above 97.5%. Lastly, we measured the computational complexity of the classifiers, in terms of central processing unit (CPU) time needed for classification, to provide a rough comparison between the algorithms in terms of battery usage requirements. As a result, the classifiers can be ranked from lowest to highest complexity (i.e., computational cost) as follows: SVM, ANN, LR, BN, DT, NB, IBk, LWL and KStar. The instance-based classifiers take considerably more computational time than the non-instance-based classifiers, whereas the slowest non-instance-based classifier (NB) required about five-times the amount of CPU time as the fastest classifier (SVM). The above results suggest that DT algorithms are excellent candidates for detecting mobility contexts in smartphones, both in terms of performance and computational complexity. PMID:25928060

Effect of sample size on multi-parametric prediction of tissue outcome in acute ischemic stroke using a random forest classifier

NASA Astrophysics Data System (ADS)

Forkert, Nils Daniel; Fiehler, Jens

2015-03-01

The tissue outcome prediction in acute ischemic stroke patients is highly relevant for clinical and research purposes. It has been shown that the combined analysis of diffusion and perfusion MRI datasets using high-level machine learning techniques leads to an improved prediction of final infarction compared to single perfusion parameter thresholding. However, most high-level classifiers require a previous training and, until now, it is ambiguous how many subjects are required for this, which is the focus of this work. 23 MRI datasets of acute stroke patients with known tissue outcome were used in this work. Relative values of diffusion and perfusion parameters as well as the binary tissue outcome were extracted on a voxel-by- voxel level for all patients and used for training of a random forest classifier. The number of patients used for training set definition was iteratively and randomly reduced from using all 22 other patients to only one other patient. Thus, 22 tissue outcome predictions were generated for each patient using the trained random forest classifiers and compared to the known tissue outcome using the Dice coefficient. Overall, a logarithmic relation between the number of patients used for training set definition and tissue outcome prediction accuracy was found. Quantitatively, a mean Dice coefficient of 0.45 was found for the prediction using the training set consisting of the voxel information from only one other patient, which increases to 0.53 if using all other patients (n=22). Based on extrapolation, 50-100 patients appear to be a reasonable tradeoff between tissue outcome prediction accuracy and effort required for data acquisition and preparation.

Automatic vetting of planet candidates from ground based surveys: Machine learning with NGTS

NASA Astrophysics Data System (ADS)

Armstrong, David J.; Günther, Maximilian N.; McCormac, James; Smith, Alexis M. S.; Bayliss, Daniel; Bouchy, François; Burleigh, Matthew R.; Casewell, Sarah; Eigmüller, Philipp; Gillen, Edward; Goad, Michael R.; Hodgkin, Simon T.; Jenkins, James S.; Louden, Tom; Metrailler, Lionel; Pollacco, Don; Poppenhaeger, Katja; Queloz, Didier; Raynard, Liam; Rauer, Heike; Udry, Stéphane; Walker, Simon R.; Watson, Christopher A.; West, Richard G.; Wheatley, Peter J.

2018-05-01

State of the art exoplanet transit surveys are producing ever increasing quantities of data. To make the best use of this resource, in detecting interesting planetary systems or in determining accurate planetary population statistics, requires new automated methods. Here we describe a machine learning algorithm that forms an integral part of the pipeline for the NGTS transit survey, demonstrating the efficacy of machine learning in selecting planetary candidates from multi-night ground based survey data. Our method uses a combination of random forests and self-organising-maps to rank planetary candidates, achieving an AUC score of 97.6% in ranking 12368 injected planets against 27496 false positives in the NGTS data. We build on past examples by using injected transit signals to form a training set, a necessary development for applying similar methods to upcoming surveys. We also make the autovet code used to implement the algorithm publicly accessible. autovet is designed to perform machine learned vetting of planetary candidates, and can utilise a variety of methods. The apparent robustness of machine learning techniques, whether on space-based or the qualitatively different ground-based data, highlights their importance to future surveys such as TESS and PLATO and the need to better understand their advantages and pitfalls in an exoplanetary context.

Risk estimation using probability machines

PubMed Central

2014-01-01

Background Logistic regression has been the de facto, and often the only, model used in the description and analysis of relationships between a binary outcome and observed features. It is widely used to obtain the conditional probabilities of the outcome given predictors, as well as predictor effect size estimates using conditional odds ratios. Results We show how statistical learning machines for binary outcomes, provably consistent for the nonparametric regression problem, can be used to provide both consistent conditional probability estimation and conditional effect size estimates. Effect size estimates from learning machines leverage our understanding of counterfactual arguments central to the interpretation of such estimates. We show that, if the data generating model is logistic, we can recover accurate probability predictions and effect size estimates with nearly the same efficiency as a correct logistic model, both for main effects and interactions. We also propose a method using learning machines to scan for possible interaction effects quickly and efficiently. Simulations using random forest probability machines are presented. Conclusions The models we propose make no assumptions about the data structure, and capture the patterns in the data by just specifying the predictors involved and not any particular model structure. So they do not run the same risks of model mis-specification and the resultant estimation biases as a logistic model. This methodology, which we call a “risk machine”, will share properties from the statistical machine that it is derived from. PMID:24581306

Orchid: a novel management, annotation and machine learning framework for analyzing cancer mutations.

PubMed

Cario, Clinton L; Witte, John S

2018-03-15

As whole-genome tumor sequence and biological annotation datasets grow in size, number and content, there is an increasing basic science and clinical need for efficient and accurate data management and analysis software. With the emergence of increasingly sophisticated data stores, execution environments and machine learning algorithms, there is also a need for the integration of functionality across frameworks. We present orchid, a python based software package for the management, annotation and machine learning of cancer mutations. Building on technologies of parallel workflow execution, in-memory database storage and machine learning analytics, orchid efficiently handles millions of mutations and hundreds of features in an easy-to-use manner. We describe the implementation of orchid and demonstrate its ability to distinguish tissue of origin in 12 tumor types based on 339 features using a random forest classifier. Orchid and our annotated tumor mutation database are freely available at https://github.com/wittelab/orchid. Software is implemented in python 2.7, and makes use of MySQL or MemSQL databases. Groovy 2.4.5 is optionally required for parallel workflow execution. JWitte@ucsf.edu. Supplementary data are available at Bioinformatics online.

Identifying Green Infrastructure from Social Media and Crowdsourcing- An Image Based Machine-Learning Approach.

NASA Astrophysics Data System (ADS)

Rai, A.; Minsker, B. S.

2016-12-01

In this work we introduce a novel dataset GRID: GReen Infrastructure Detection Dataset and a framework for identifying urban green storm water infrastructure (GI) designs (wetlands/ponds, urban trees, and rain gardens/bioswales) from social media and satellite aerial images using computer vision and machine learning methods. Along with the hydrologic benefits of GI, such as reducing runoff volumes and urban heat islands, GI also provides important socio-economic benefits such as stress recovery and community cohesion. However, GI is installed by many different parties and cities typically do not know where GI is located, making study of its impacts or siting new GI difficult. We use object recognition learning methods (template matching, sliding window approach, and Random Hough Forest method) and supervised machine learning algorithms (e.g., support vector machines) as initial screening approaches to detect potential GI sites, which can then be investigated in more detail using on-site surveys. Training data were collected from GPS locations of Flickr and Instagram image postings and Amazon Mechanical Turk identification of each GI type. Sliding window method outperformed other methods and achieved an average F measure, which is combined metric for precision and recall performance measure of 0.78.

Ecological interactions and the Netflix problem.

PubMed

Desjardins-Proulx, Philippe; Laigle, Idaline; Poisot, Timothée; Gravel, Dominique

2017-01-01

Species interactions are a key component of ecosystems but we generally have an incomplete picture of who-eats-who in a given community. Different techniques have been devised to predict species interactions using theoretical models or abundances. Here, we explore the K nearest neighbour approach, with a special emphasis on recommendation, along with a supervised machine learning technique. Recommenders are algorithms developed for companies like Netflix to predict whether a customer will like a product given the preferences of similar customers. These machine learning techniques are well-suited to study binary ecological interactions since they focus on positive-only data. By removing a prey from a predator, we find that recommenders can guess the missing prey around 50% of the times on the first try, with up to 881 possibilities. Traits do not improve significantly the results for the K nearest neighbour, although a simple test with a supervised learning approach (random forests) show we can predict interactions with high accuracy using only three traits per species. This result shows that binary interactions can be predicted without regard to the ecological community given only three variables: body mass and two variables for the species' phylogeny. These techniques are complementary, as recommenders can predict interactions in the absence of traits, using only information about other species' interactions, while supervised learning algorithms such as random forests base their predictions on traits only but do not exploit other species' interactions. Further work should focus on developing custom similarity measures specialized for ecology to improve the KNN algorithms and using richer data to capture indirect relationships between species.

Ecological interactions and the Netflix problem

PubMed Central

Laigle, Idaline; Poisot, Timothée; Gravel, Dominique

2017-01-01

Species interactions are a key component of ecosystems but we generally have an incomplete picture of who-eats-who in a given community. Different techniques have been devised to predict species interactions using theoretical models or abundances. Here, we explore the K nearest neighbour approach, with a special emphasis on recommendation, along with a supervised machine learning technique. Recommenders are algorithms developed for companies like Netflix to predict whether a customer will like a product given the preferences of similar customers. These machine learning techniques are well-suited to study binary ecological interactions since they focus on positive-only data. By removing a prey from a predator, we find that recommenders can guess the missing prey around 50% of the times on the first try, with up to 881 possibilities. Traits do not improve significantly the results for the K nearest neighbour, although a simple test with a supervised learning approach (random forests) show we can predict interactions with high accuracy using only three traits per species. This result shows that binary interactions can be predicted without regard to the ecological community given only three variables: body mass and two variables for the species’ phylogeny. These techniques are complementary, as recommenders can predict interactions in the absence of traits, using only information about other species’ interactions, while supervised learning algorithms such as random forests base their predictions on traits only but do not exploit other species’ interactions. Further work should focus on developing custom similarity measures specialized for ecology to improve the KNN algorithms and using richer data to capture indirect relationships between species. PMID:28828250

Sentinel node status prediction by four statistical models: results from a large bi-institutional series (n = 1132).

PubMed

Mocellin, Simone; Thompson, John F; Pasquali, Sandro; Montesco, Maria C; Pilati, Pierluigi; Nitti, Donato; Saw, Robyn P; Scolyer, Richard A; Stretch, Jonathan R; Rossi, Carlo R

2009-12-01

To improve selection for sentinel node (SN) biopsy (SNB) in patients with cutaneous melanoma using statistical models predicting SN status. About 80% of patients currently undergoing SNB are node negative. In the absence of conclusive evidence of a SNBassociated survival benefit, these patients may be over-treated. Here, we tested the efficiency of 4 different models in predicting SN status. The clinicopathologic data (age, gender, tumor thickness, Clark level, regression, ulceration, histologic subtype, and mitotic index) of 1132 melanoma patients who had undergone SNB at institutions in Italy and Australia were analyzed. Logistic regression, classification tree, random forest, and support vector machine models were fitted to the data. The predictive models were built with the aim of maximizing the negative predictive value (NPV) and reducing the rate of SNB procedures though minimizing the error rate. After cross-validation logistic regression, classification tree, random forest, and support vector machine predictive models obtained clinically relevant NPV (93.6%, 94.0%, 97.1%, and 93.0%, respectively), SNB reduction (27.5%, 29.8%, 18.2%, and 30.1%, respectively), and error rates (1.8%, 1.8%, 0.5%, and 2.1%, respectively). Using commonly available clinicopathologic variables, predictive models can preoperatively identify a proportion of patients ( approximately 25%) who might be spared SNB, with an acceptable (1%-2%) error. If validated in large prospective series, these models might be implemented in the clinical setting for improved patient selection, which ultimately would lead to better quality of life for patients and optimization of resource allocation for the health care system.

Development of an automated assessment tool for MedWatch reports in the FDA adverse event reporting system.

PubMed

Han, Lichy; Ball, Robert; Pamer, Carol A; Altman, Russ B; Proestel, Scott

2017-09-01

As the US Food and Drug Administration (FDA) receives over a million adverse event reports associated with medication use every year, a system is needed to aid FDA safety evaluators in identifying reports most likely to demonstrate causal relationships to the suspect medications. We combined text mining with machine learning to construct and evaluate such a system to identify medication-related adverse event reports. FDA safety evaluators assessed 326 reports for medication-related causality. We engineered features from these reports and constructed random forest, L1 regularized logistic regression, and support vector machine models. We evaluated model accuracy and further assessed utility by generating report rankings that represented a prioritized report review process. Our random forest model showed the best performance in report ranking and accuracy, with an area under the receiver operating characteristic curve of 0.66. The generated report ordering assigns reports with a higher probability of medication-related causality a higher rank and is significantly correlated to a perfect report ordering, with a Kendall's tau of 0.24 ( P  = .002). Our models produced prioritized report orderings that enable FDA safety evaluators to focus on reports that are more likely to contain valuable medication-related adverse event information. Applying our models to all FDA adverse event reports has the potential to streamline the manual review process and greatly reduce reviewer workload. Published by Oxford University Press on behalf of the American Medical Informatics Association 2017. This work is written by US Government employees and is in the public domain in the United States.

Validation of optical codes based on 3D nanostructures

NASA Astrophysics Data System (ADS)

Carnicer, Artur; Javidi, Bahram

2017-05-01

Image information encoding using random phase masks produce speckle-like noise distributions when the sample is propagated in the Fresnel domain. As a result, information cannot be accessed by simple visual inspection. Phase masks can be easily implemented in practice by attaching cello-tape to the plain-text message. Conventional 2D-phase masks can be generalized to 3D by combining glass and diffusers resulting in a more complex, physical unclonable function. In this communication, we model the behavior of a 3D phase mask using a simple approach: light is propagated trough glass using the angular spectrum of plane waves whereas the diffusor is described as a random phase mask and a blurring effect on the amplitude of the propagated wave. Using different designs for the 3D phase mask and multiple samples, we demonstrate that classification is possible using the k-nearest neighbors and random forests machine learning algorithms.

A machine learning calibration model using random forests to improve sensor performance for lower-cost air quality monitoring

NASA Astrophysics Data System (ADS)

Zimmerman, Naomi; Presto, Albert A.; Kumar, Sriniwasa P. N.; Gu, Jason; Hauryliuk, Aliaksei; Robinson, Ellis S.; Robinson, Allen L.; Subramanian, R.

2018-01-01

Low-cost sensing strategies hold the promise of denser air quality monitoring networks, which could significantly improve our understanding of personal air pollution exposure. Additionally, low-cost air quality sensors could be deployed to areas where limited monitoring exists. However, low-cost sensors are frequently sensitive to environmental conditions and pollutant cross-sensitivities, which have historically been poorly addressed by laboratory calibrations, limiting their utility for monitoring. In this study, we investigated different calibration models for the Real-time Affordable Multi-Pollutant (RAMP) sensor package, which measures CO, NO2, O3, and CO2. We explored three methods: (1) laboratory univariate linear regression, (2) empirical multiple linear regression, and (3) machine-learning-based calibration models using random forests (RF). Calibration models were developed for 16-19 RAMP monitors (varied by pollutant) using training and testing windows spanning August 2016 through February 2017 in Pittsburgh, PA, US. The random forest models matched (CO) or significantly outperformed (NO2, CO2, O3) the other calibration models, and their accuracy and precision were robust over time for testing windows of up to 16 weeks. Following calibration, average mean absolute error on the testing data set from the random forest models was 38 ppb for CO (14 % relative error), 10 ppm for CO2 (2 % relative error), 3.5 ppb for NO2 (29 % relative error), and 3.4 ppb for O3 (15 % relative error), and Pearson r versus the reference monitors exceeded 0.8 for most units. Model performance is explored in detail, including a quantification of model variable importance, accuracy across different concentration ranges, and performance in a range of monitoring contexts including the National Ambient Air Quality Standards (NAAQS) and the US EPA Air Sensors Guidebook recommendations of minimum data quality for personal exposure measurement. A key strength of the RF approach is that it accounts for pollutant cross-sensitivities. This highlights the importance of developing multipollutant sensor packages (as opposed to single-pollutant monitors); we determined this is especially critical for NO2 and CO2. The evaluation reveals that only the RF-calibrated sensors meet the US EPA Air Sensors Guidebook recommendations of minimum data quality for personal exposure measurement. We also demonstrate that the RF-model-calibrated sensors could detect differences in NO2 concentrations between a near-road site and a suburban site less than 1.5 km away. From this study, we conclude that combining RF models with carefully controlled state-of-the-art multipollutant sensor packages as in the RAMP monitors appears to be a very promising approach to address the poor performance that has plagued low-cost air quality sensors.

Utilization and cost for animal logging operations

Treesearch

Suraj P. Shrestha; Bobby L. Lanford

2001-01-01

Forest harvesting with animals is a labor-intensive operation. Due to the development of efficient machines and high volume demands from the forest products industry, mechanization of logging developed very fast, leaving behind the traditional horse and mule logging. It is expensive to use machines on smaller woodlots, which require frequent moves if mechanically...

Intra-and-Inter Species Biomass Prediction in a Plantation Forest: Testing the Utility of High Spatial Resolution Spaceborne Multispectral RapidEye Sensor and Advanced Machine Learning Algorithms

PubMed Central

Dube, Timothy; Mutanga, Onisimo; Adam, Elhadi; Ismail, Riyad

2014-01-01

The quantification of aboveground biomass using remote sensing is critical for better understanding the role of forests in carbon sequestration and for informed sustainable management. Although remote sensing techniques have been proven useful in assessing forest biomass in general, more is required to investigate their capabilities in predicting intra-and-inter species biomass which are mainly characterised by non-linear relationships. In this study, we tested two machine learning algorithms, Stochastic Gradient Boosting (SGB) and Random Forest (RF) regression trees to predict intra-and-inter species biomass using high resolution RapidEye reflectance bands as well as the derived vegetation indices in a commercial plantation. The results showed that the SGB algorithm yielded the best performance for intra-and-inter species biomass prediction; using all the predictor variables as well as based on the most important selected variables. For example using the most important variables the algorithm produced an R2 of 0.80 and RMSE of 16.93 t·ha−1 for E. grandis; R2 of 0.79, RMSE of 17.27 t·ha−1 for P. taeda and R2 of 0.61, RMSE of 43.39 t·ha−1 for the combined species data sets. Comparatively, RF yielded plausible results only for E. dunii (R2 of 0.79; RMSE of 7.18 t·ha−1). We demonstrated that although the two statistical methods were able to predict biomass accurately, RF produced weaker results as compared to SGB when applied to combined species dataset. The result underscores the relevance of stochastic models in predicting biomass drawn from different species and genera using the new generation high resolution RapidEye sensor with strategically positioned bands. PMID:25140631

Multivariate classification with random forests for gravitational wave searches of black hole binary coalescence

NASA Astrophysics Data System (ADS)

Baker, Paul T.; Caudill, Sarah; Hodge, Kari A.; Talukder, Dipongkar; Capano, Collin; Cornish, Neil J.

2015-03-01

Searches for gravitational waves produced by coalescing black hole binaries with total masses ≳25 M⊙ use matched filtering with templates of short duration. Non-Gaussian noise bursts in gravitational wave detector data can mimic short signals and limit the sensitivity of these searches. Previous searches have relied on empirically designed statistics incorporating signal-to-noise ratio and signal-based vetoes to separate gravitational wave candidates from noise candidates. We report on sensitivity improvements achieved using a multivariate candidate ranking statistic derived from a supervised machine learning algorithm. We apply the random forest of bagged decision trees technique to two separate searches in the high mass (≳25 M⊙ ) parameter space. For a search which is sensitive to gravitational waves from the inspiral, merger, and ringdown of binary black holes with total mass between 25 M⊙ and 100 M⊙ , we find sensitive volume improvements as high as 70±13%-109±11% when compared to the previously used ranking statistic. For a ringdown-only search which is sensitive to gravitational waves from the resultant perturbed intermediate mass black hole with mass roughly between 10 M⊙ and 600 M⊙ , we find sensitive volume improvements as high as 61±4%-241±12% when compared to the previously used ranking statistic. We also report how sensitivity improvements can differ depending on mass regime, mass ratio, and available data quality information. Finally, we describe the techniques used to tune and train the random forest classifier that can be generalized to its use in other searches for gravitational waves.

Fangorn Forest (F2): a machine learning approach to classify genes and genera in the family Geminiviridae.

PubMed

Silva, José Cleydson F; Carvalho, Thales F M; Fontes, Elizabeth P B; Cerqueira, Fabio R

2017-09-30

Geminiviruses infect a broad range of cultivated and non-cultivated plants, causing significant economic losses worldwide. The studies of the diversity of species, taxonomy, mechanisms of evolution, geographic distribution, and mechanisms of interaction of these pathogens with the host have greatly increased in recent years. Furthermore, the use of rolling circle amplification (RCA) and advanced metagenomics approaches have enabled the elucidation of viromes and the identification of many viral agents in a large number of plant species. As a result, determining the nomenclature and taxonomically classifying geminiviruses turned into complex tasks. In addition, the gene responsible for viral replication (particularly, the viruses belonging to the genus Mastrevirus) may be spliced due to the use of the transcriptional/splicing machinery in the host cells. However, the current tools have limitations concerning the identification of introns. This study proposes a new method, designated Fangorn Forest (F2), based on machine learning approaches to classify genera using an ab initio approach, i.e., using only the genomic sequence, as well as to predict and classify genes in the family Geminiviridae. In this investigation, nine genera of the family Geminiviridae and their related satellite DNAs were selected. We obtained two training sets, one for genus classification, containing attributes extracted from the complete genome of geminiviruses, while the other was made up to classify geminivirus genes, containing attributes extracted from ORFs taken from the complete genomes cited above. Three ML algorithms were applied on those datasets to build the predictive models: support vector machines, using the sequential minimal optimization training approach, random forest (RF), and multilayer perceptron. RF demonstrated a very high predictive power, achieving 0.966, 0.964, and 0.995 of precision, recall, and area under the curve (AUC), respectively, for genus classification. For gene classification, RF could reach 0.983, 0.983, and 0.998 of precision, recall, and AUC, respectively. Therefore, Fangorn Forest is proven to be an efficient method for classifying genera of the family Geminiviridae with high precision and effective gene prediction and classification. The method is freely accessible at www.geminivirus.org:8080/geminivirusdw/discoveryGeminivirus.jsp .

Modeling long-term suspended-sediment export from an undisturbed forest catchment

NASA Astrophysics Data System (ADS)

Zimmermann, Alexander; Francke, Till; Elsenbeer, Helmut

2013-04-01

Most estimates of suspended sediment yields from humid, undisturbed, and geologically stable forest environments fall within a range of 5 - 30 t km-2 a-1. These low natural erosion rates in small headwater catchments (≤ 1 km2) support the common impression that a well-developed forest cover prevents surface erosion. Interestingly, those estimates originate exclusively from areas with prevailing vertical hydrological flow paths. Forest environments dominated by (near-) surface flow paths (overland flow, pipe flow, and return flow) and a fast response to rainfall, however, are not an exceptional phenomenon, yet only very few sediment yields have been estimated for these areas. Not surprisingly, even fewer long-term (≥ 10 years) records exist. In this contribution we present our latest research which aims at quantifying long-term suspended-sediment export from an undisturbed rainforest catchment prone to frequent overland flow. A key aspect of our approach is the application of machine-learning techniques (Random Forest, Quantile Regression Forest) which allows not only the handling of non-Gaussian data, non-linear relations between predictors and response, and correlations between predictors, but also the assessment of prediction uncertainty. For the current study we provided the machine-learning algorithms exclusively with information from a high-resolution rainfall time series to reconstruct discharge and suspended sediment dynamics for a 21-year period. The significance of our results is threefold. First, our estimates clearly show that forest cover does not necessarily prevent erosion if wet antecedent conditions and large rainfalls coincide. During these situations, overland flow is widespread and sediment fluxes increase in a non-linear fashion due to the mobilization of new sediment sources. Second, our estimates indicate that annual suspended sediment yields of the undisturbed forest catchment show large fluctuations. Depending on the frequency of large events, annual suspended-sediment yield varies between 74 - 416 t km-2 a-1. Third, the estimated sediment yields exceed former benchmark values by an order of magnitude and provide evidence that the erosion footprint of undisturbed, forested catchments can be undistinguishable from that of sustainably managed, but hydrologically less responsive areas. Because of the susceptibility to soil loss we argue that any land use should be avoided in natural erosion hotspots.

Novel Breast Imaging and Machine Learning: Predicting Breast Lesion Malignancy at Cone-Beam CT Using Machine Learning Techniques.

PubMed

Uhlig, Johannes; Uhlig, Annemarie; Kunze, Meike; Beissbarth, Tim; Fischer, Uwe; Lotz, Joachim; Wienbeck, Susanne

2018-05-24

The purpose of this study is to evaluate the diagnostic performance of machine learning techniques for malignancy prediction at breast cone-beam CT (CBCT) and to compare them to human readers. Five machine learning techniques, including random forests, back propagation neural networks (BPN), extreme learning machines, support vector machines, and K-nearest neighbors, were used to train diagnostic models on a clinical breast CBCT dataset with internal validation by repeated 10-fold cross-validation. Two independent blinded human readers with profound experience in breast imaging and breast CBCT analyzed the same CBCT dataset. Diagnostic performance was compared using AUC, sensitivity, and specificity. The clinical dataset comprised 35 patients (American College of Radiology density type C and D breasts) with 81 suspicious breast lesions examined with contrast-enhanced breast CBCT. Forty-five lesions were histopathologically proven to be malignant. Among the machine learning techniques, BPNs provided the best diagnostic performance, with AUC of 0.91, sensitivity of 0.85, and specificity of 0.82. The diagnostic performance of the human readers was AUC of 0.84, sensitivity of 0.89, and specificity of 0.72 for reader 1 and AUC of 0.72, sensitivity of 0.71, and specificity of 0.67 for reader 2. AUC was significantly higher for BPN when compared with both reader 1 (p = 0.01) and reader 2 (p < 0.001). Machine learning techniques provide a high and robust diagnostic performance in the prediction of malignancy in breast lesions identified at CBCT. BPNs showed the best diagnostic performance, surpassing human readers in terms of AUC and specificity.

Shaping-lathe headrig will stretch shrinking timber supply

Treesearch

J. Gengler; J.D. Saul

1975-01-01

The first commercial version of the shaping lathe headrig, designed to machine short hardwood or softwood logs into cants and flakes, was introduced to forest industry executives in September during a working demonstration at Stetson-Ross Machine Co., Seattle. Based on a concept provided by Dr. Peter Koch, chief wood scientist at the Southern Forest Experiment Station...

Machine Vision Technology for the Forest Products Industry

Treesearch

Richard W. Conners; D.Earl Kline; Philip A. Araman; Thomas T. Drayer

1997-01-01

From forest to finished product, wood is moved from one processing stage to the next, subject to the decisions of individuals along the way. While this process has worked for hundreds of years, the technology exists today to provide more complete information to the decision makers. Virginia Tech has developed this technology, creating a machine vision prototype for...

Using Machine Learning To Predict Which Light Curves Will Yield Stellar Rotation Periods

NASA Astrophysics Data System (ADS)

Agüeros, Marcel; Teachey, Alexander

2018-01-01

Using time-domain photometry to reliably measure a solar-type star's rotation period requires that its light curve have a number of favorable characteristics. The probability of recovering a period will be a non-linear function of these light curve features, which are either astrophysical in nature or set by the observations. We employ standard machine learning algorithms (artificial neural networks and random forests) to predict whether a given light curve will produce a robust rotation period measurement from its Lomb-Scargle periodogram. The algorithms are trained and validated using salient statistics extracted from both simulated light curves and their corresponding periodograms, and we apply these classifiers to the most recent Intermediate Palomar Transient Factory (iPTF) data release. With this pipeline, we anticipate measuring rotation periods for a significant fraction of the ∼4x108 stars in the iPTF footprint.

Unsteady flow sensing and optimal sensor placement using machine learning

NASA Astrophysics Data System (ADS)

Semaan, Richard

2016-11-01

Machine learning is used to estimate the flow state and to determine the optimal sensor placement over a two-dimensional (2D) airfoil equipped with a Coanda actuator. The analysis is based on flow field data obtained from 2D unsteady Reynolds averaged Navier-Stokes (uRANS) simulations with different jet blowing intensities and actuation frequencies, characterizing different flow separation states. This study shows how the "random forests" algorithm is utilized beyond its typical usage in fluid mechanics estimating the flow state to determine the optimal sensor placement. The results are compared against the current de-facto standard of maximum modal amplitude location and against a brute force approach that scans all possible sensor combinations. The results show that it is possible to simultaneously infer the state of flow and to determine the optimal sensor location without the need to perform proper orthogonal decomposition. Collaborative Research Center (CRC) 880, DFG.

Cosmic string detection with tree-based machine learning

NASA Astrophysics Data System (ADS)

Vafaei Sadr, A.; Farhang, M.; Movahed, S. M. S.; Bassett, B.; Kunz, M.

2018-07-01

We explore the use of random forest and gradient boosting, two powerful tree-based machine learning algorithms, for the detection of cosmic strings in maps of the cosmic microwave background (CMB), through their unique Gott-Kaiser-Stebbins effect on the temperature anisotropies. The information in the maps is compressed into feature vectors before being passed to the learning units. The feature vectors contain various statistical measures of the processed CMB maps that boost cosmic string detectability. Our proposed classifiers, after training, give results similar to or better than claimed detectability levels from other methods for string tension, Gμ. They can make 3σ detection of strings with Gμ ≳ 2.1 × 10-10 for noise-free, 0.9'-resolution CMB observations. The minimum detectable tension increases to Gμ ≳ 3.0 × 10-8 for a more realistic, CMB S4-like (II) strategy, improving over previous results.

Cosmic String Detection with Tree-Based Machine Learning

NASA Astrophysics Data System (ADS)

Vafaei Sadr, A.; Farhang, M.; Movahed, S. M. S.; Bassett, B.; Kunz, M.

2018-05-01

We explore the use of random forest and gradient boosting, two powerful tree-based machine learning algorithms, for the detection of cosmic strings in maps of the cosmic microwave background (CMB), through their unique Gott-Kaiser-Stebbins effect on the temperature anisotropies. The information in the maps is compressed into feature vectors before being passed to the learning units. The feature vectors contain various statistical measures of the processed CMB maps that boost cosmic string detectability. Our proposed classifiers, after training, give results similar to or better than claimed detectability levels from other methods for string tension, Gμ. They can make 3σ detection of strings with Gμ ≳ 2.1 × 10-10 for noise-free, 0.9΄-resolution CMB observations. The minimum detectable tension increases to Gμ ≳ 3.0 × 10-8 for a more realistic, CMB S4-like (II) strategy, improving over previous results.

Interpreting Black-Box Classifiers Using Instance-Level Visual Explanations

DOE Office of Scientific and Technical Information (OSTI.GOV)

Tamagnini, Paolo; Krause, Josua W.; Dasgupta, Aritra

2017-05-14

To realize the full potential of machine learning in diverse real- world domains, it is necessary for model predictions to be readily interpretable and actionable for the human in the loop. Analysts, who are the users but not the developers of machine learning models, often do not trust a model because of the lack of transparency in associating predictions with the underlying data space. To address this problem, we propose Rivelo, a visual analytic interface that enables analysts to understand the causes behind predictions of binary classifiers by interactively exploring a set of instance-level explanations. These explanations are model-agnostic, treatingmore » a model as a black box, and they help analysts in interactively probing the high-dimensional binary data space for detecting features relevant to predictions. We demonstrate the utility of the interface with a case study analyzing a random forest model on the sentiment of Yelp reviews about doctors.« less

Biochemical Profile of Heritage and Modern Apple Cultivars and Application of Machine Learning Methods To Predict Usage, Age, and Harvest Season.

PubMed

Anastasiadi, Maria; Mohareb, Fady; Redfern, Sally P; Berry, Mark; Simmonds, Monique S J; Terry, Leon A

2017-07-05

The present study represents the first major attempt to characterize the biochemical profile in different tissues of a large selection of apple cultivars sourced from the United Kingdom's National Fruit Collection comprising dessert, ornamental, cider, and culinary apples. Furthermore, advanced machine learning methods were applied with the objective to identify whether the phenolic and sugar composition of an apple cultivar could be used as a biomarker fingerprint to differentiate between heritage and mainstream commercial cultivars as well as govern the separation among primary usage groups and harvest season. A prediction accuracy of >90% was achieved with the random forest method for all three models. The results highlighted the extraordinary phytochemical potency and unique profile of some heritage, cider, and ornamental apple cultivars, especially in comparison to more mainstream apple cultivars. Therefore, these findings could guide future cultivar selection on the basis of health-promoting phytochemical content.

Applying data fusion techniques for benthic habitat mapping and monitoring in a coral reef ecosystem

NASA Astrophysics Data System (ADS)

Zhang, Caiyun

2015-06-01

Accurate mapping and effective monitoring of benthic habitat in the Florida Keys are critical in developing management strategies for this valuable coral reef ecosystem. For this study, a framework was designed for automated benthic habitat mapping by combining multiple data sources (hyperspectral, aerial photography, and bathymetry data) and four contemporary imagery processing techniques (data fusion, Object-based Image Analysis (OBIA), machine learning, and ensemble analysis). In the framework, 1-m digital aerial photograph was first merged with 17-m hyperspectral imagery and 10-m bathymetry data using a pixel/feature-level fusion strategy. The fused dataset was then preclassified by three machine learning algorithms (Random Forest, Support Vector Machines, and k-Nearest Neighbor). Final object-based habitat maps were produced through ensemble analysis of outcomes from three classifiers. The framework was tested for classifying a group-level (3-class) and code-level (9-class) habitats in a portion of the Florida Keys. Informative and accurate habitat maps were achieved with an overall accuracy of 88.5% and 83.5% for the group-level and code-level classifications, respectively.

Autonomous unobtrusive detection of mild cognitive impairment in older adults.

PubMed

Akl, Ahmad; Taati, Babak; Mihailidis, Alex

2015-05-01

The current diagnosis process of dementia is resulting in a high percentage of cases with delayed detection. To address this problem, in this paper, we explore the feasibility of autonomously detecting mild cognitive impairment (MCI) in the older adult population. We implement a signal processing approach equipped with a machine learning paradigm to process and analyze real-world data acquired using home-based unobtrusive sensing technologies. Using the sensor and clinical data pertaining to 97 subjects, acquired over an average period of three years, a number of measures associated with the subjects' walking speed and general activity in the home were calculated. Different time spans of these measures were used to generate feature vectors to train and test two machine learning algorithms namely support vector machines and random forests. We were able to autonomously detect MCI in older adults with an area under the ROC curve of 0.97 and an area under the precision-recall curve of 0.93 using a time window of 24 weeks. This study is of great significance since it can potentially assist in the early detection of cognitive impairment in older adults.

Building blocks for automated elucidation of metabolites: machine learning methods for NMR prediction.

PubMed

Kuhn, Stefan; Egert, Björn; Neumann, Steffen; Steinbeck, Christoph

2008-09-25

Current efforts in Metabolomics, such as the Human Metabolome Project, collect structures of biological metabolites as well as data for their characterisation, such as spectra for identification of substances and measurements of their concentration. Still, only a fraction of existing metabolites and their spectral fingerprints are known. Computer-Assisted Structure Elucidation (CASE) of biological metabolites will be an important tool to leverage this lack of knowledge. Indispensable for CASE are modules to predict spectra for hypothetical structures. This paper evaluates different statistical and machine learning methods to perform predictions of proton NMR spectra based on data from our open database NMRShiftDB. A mean absolute error of 0.18 ppm was achieved for the prediction of proton NMR shifts ranging from 0 to 11 ppm. Random forest, J48 decision tree and support vector machines achieved similar overall errors. HOSE codes being a notably simple method achieved a comparatively good result of 0.17 ppm mean absolute error. NMR prediction methods applied in the course of this work delivered precise predictions which can serve as a building block for Computer-Assisted Structure Elucidation for biological metabolites.

Combining Benford's Law and machine learning to detect money laundering. An actual Spanish court case.

PubMed

Badal-Valero, Elena; Alvarez-Jareño, José A; Pavía, Jose M

2018-01-01

This paper is based on the analysis of the database of operations from a macro-case on money laundering orchestrated between a core company and a group of its suppliers, 26 of which had already been identified by the police as fraudulent companies. In the face of a well-founded suspicion that more companies have perpetrated criminal acts and in order to make better use of what are very limited police resources, we aim to construct a tool to detect money laundering criminals. We combine Benford's Law and machine learning algorithms (logistic regression, decision trees, neural networks, and random forests) to find patterns of money laundering criminals in the context of a real Spanish court case. After mapping each supplier's set of accounting data into a 21-dimensional space using Benford's Law and applying machine learning algorithms, additional companies that could merit further scrutiny are flagged up. A new tool to detect money laundering criminals is proposed in this paper. The tool is tested in the context of a real case. Copyright © 2017 Elsevier B.V. All rights reserved.

Semi-supervised prediction of gene regulatory networks using machine learning algorithms.

PubMed

Patel, Nihir; Wang, Jason T L

2015-10-01

Use of computational methods to predict gene regulatory networks (GRNs) from gene expression data is a challenging task. Many studies have been conducted using unsupervised methods to fulfill the task; however, such methods usually yield low prediction accuracies due to the lack of training data. In this article, we propose semi-supervised methods for GRN prediction by utilizing two machine learning algorithms, namely, support vector machines (SVM) and random forests (RF). The semi-supervised methods make use of unlabelled data for training. We investigated inductive and transductive learning approaches, both of which adopt an iterative procedure to obtain reliable negative training data from the unlabelled data. We then applied our semi-supervised methods to gene expression data of Escherichia coli and Saccharomyces cerevisiae, and evaluated the performance of our methods using the expression data. Our analysis indicated that the transductive learning approach outperformed the inductive learning approach for both organisms. However, there was no conclusive difference identified in the performance of SVM and RF. Experimental results also showed that the proposed semi-supervised methods performed better than existing supervised methods for both organisms.

Foliar and woody materials discriminated using terrestrial LiDAR in a mixed natural forest

NASA Astrophysics Data System (ADS)

Zhu, Xi; Skidmore, Andrew K.; Darvishzadeh, Roshanak; Niemann, K. Olaf; Liu, Jing; Shi, Yifang; Wang, Tiejun

2018-02-01

Separation of foliar and woody materials using remotely sensed data is crucial for the accurate estimation of leaf area index (LAI) and woody biomass across forest stands. In this paper, we present a new method to accurately separate foliar and woody materials using terrestrial LiDAR point clouds obtained from ten test sites in a mixed forest in Bavarian Forest National Park, Germany. Firstly, we applied and compared an adaptive radius near-neighbor search algorithm with a fixed radius near-neighbor search method in order to obtain both radiometric and geometric features derived from terrestrial LiDAR point clouds. Secondly, we used a random forest machine learning algorithm to classify foliar and woody materials and examined the impact of understory and slope on the classification accuracy. An average overall accuracy of 84.4% (Kappa = 0.75) was achieved across all experimental plots. The adaptive radius near-neighbor search method outperformed the fixed radius near-neighbor search method. The classification accuracy was significantly higher when the combination of both radiometric and geometric features was utilized. The analysis showed that increasing slope and understory coverage had a significant negative effect on the overall classification accuracy. Our results suggest that the utilization of the adaptive radius near-neighbor search method coupling both radiometric and geometric features has the potential to accurately discriminate foliar and woody materials from terrestrial LiDAR data in a mixed natural forest.

Machine Learning Algorithms for prediction of regions of high Reynolds Averaged Navier Stokes Uncertainty

NASA Astrophysics Data System (ADS)

Mishra, Aashwin; Iaccarino, Gianluca

2017-11-01

In spite of their deficiencies, RANS models represent the workhorse for industrial investigations into turbulent flows. In this context, it is essential to provide diagnostic measures to assess the quality of RANS predictions. To this end, the primary step is to identify feature importances amongst massive sets of potentially descriptive and discriminative flow features. This aids the physical interpretability of the resultant discrepancy model and its extensibility to similar problems. Recent investigations have utilized approaches such as Random Forests, Support Vector Machines and the Least Absolute Shrinkage and Selection Operator for feature selection. With examples, we exhibit how such methods may not be suitable for turbulent flow datasets. The underlying rationale, such as the correlation bias and the required conditions for the success of penalized algorithms, are discussed with illustrative examples. Finally, we provide alternate approaches using convex combinations of regularized regression approaches and randomized sub-sampling in combination with feature selection algorithms, to infer model structure from data. This research was supported by the Defense Advanced Research Projects Agency under the Enabling Quantification of Uncertainty in Physical Systems (EQUiPS) project (technical monitor: Dr Fariba Fahroo).

Semi-supervised manifold learning with affinity regularization for Alzheimer's disease identification using positron emission tomography imaging.

PubMed

Lu, Shen; Xia, Yong; Cai, Tom Weidong; Feng, David Dagan

2015-01-01

Dementia, Alzheimer's disease (AD) in particular is a global problem and big threat to the aging population. An image based computer-aided dementia diagnosis method is needed to providing doctors help during medical image examination. Many machine learning based dementia classification methods using medical imaging have been proposed and most of them achieve accurate results. However, most of these methods make use of supervised learning requiring fully labeled image dataset, which usually is not practical in real clinical environment. Using large amount of unlabeled images can improve the dementia classification performance. In this study we propose a new semi-supervised dementia classification method based on random manifold learning with affinity regularization. Three groups of spatial features are extracted from positron emission tomography (PET) images to construct an unsupervised random forest which is then used to regularize the manifold learning objective function. The proposed method, stat-of-the-art Laplacian support vector machine (LapSVM) and supervised SVM are applied to classify AD and normal controls (NC). The experiment results show that learning with unlabeled images indeed improves the classification performance. And our method outperforms LapSVM on the same dataset.

Automated Identification of Abnormal Adult EEGs

PubMed Central

López, S.; Suarez, G.; Jungreis, D.; Obeid, I.; Picone, J.

2016-01-01

The interpretation of electroencephalograms (EEGs) is a process that is still dependent on the subjective analysis of the examiners. Though interrater agreement on critical events such as seizures is high, it is much lower on subtler events (e.g., when there are benign variants). The process used by an expert to interpret an EEG is quite subjective and hard to replicate by machine. The performance of machine learning technology is far from human performance. We have been developing an interpretation system, AutoEEG, with a goal of exceeding human performance on this task. In this work, we are focusing on one of the early decisions made in this process – whether an EEG is normal or abnormal. We explore two baseline classification algorithms: k-Nearest Neighbor (kNN) and Random Forest Ensemble Learning (RF). A subset of the TUH EEG Corpus was used to evaluate performance. Principal Components Analysis (PCA) was used to reduce the dimensionality of the data. kNN achieved a 41.8% detection error rate while RF achieved an error rate of 31.7%. These error rates are significantly lower than those obtained by random guessing based on priors (49.5%). The majority of the errors were related to misclassification of normal EEGs. PMID:27195311

A comparison of machine learning and Bayesian modelling for molecular serotyping.

PubMed

Newton, Richard; Wernisch, Lorenz

2017-08-11

Streptococcus pneumoniae is a human pathogen that is a major cause of infant mortality. Identifying the pneumococcal serotype is an important step in monitoring the impact of vaccines used to protect against disease. Genomic microarrays provide an effective method for molecular serotyping. Previously we developed an empirical Bayesian model for the classification of serotypes from a molecular serotyping array. With only few samples available, a model driven approach was the only option. In the meanwhile, several thousand samples have been made available to us, providing an opportunity to investigate serotype classification by machine learning methods, which could complement the Bayesian model. We compare the performance of the original Bayesian model with two machine learning algorithms: Gradient Boosting Machines and Random Forests. We present our results as an example of a generic strategy whereby a preliminary probabilistic model is complemented or replaced by a machine learning classifier once enough data are available. Despite the availability of thousands of serotyping arrays, a problem encountered when applying machine learning methods is the lack of training data containing mixtures of serotypes; due to the large number of possible combinations. Most of the available training data comprises samples with only a single serotype. To overcome the lack of training data we implemented an iterative analysis, creating artificial training data of serotype mixtures by combining raw data from single serotype arrays. With the enhanced training set the machine learning algorithms out perform the original Bayesian model. However, for serotypes currently lacking sufficient training data the best performing implementation was a combination of the results of the Bayesian Model and the Gradient Boosting Machine. As well as being an effective method for classifying biological data, machine learning can also be used as an efficient method for revealing subtle biological insights, which we illustrate with an example.

Neck pain combined with arm pain among professional drivers of forest machines and the association with whole-body vibration exposure.

PubMed

Rehn, B; Nilsson, T; Lundström, R; Hagberg, M; Burström, L

2009-10-01

The purpose of this study was to investigate the existence of neck pain and arm pain among professional forest machine drivers and to find out if pain were related to their whole-body vibration (WBV) exposure. A self-administered questionnaire was sent to 529 forest machine drivers in northern Sweden and the response was 63%. Two pain groups were formed; 1) neck pain; 2) neck pain combined with arm pain. From WBV exposure data (recent measurements made according to ISO 2631-1, available information from reports) and from the self-administered questionnaire, 14 various WBV exposure/dose measures were calculated for each driver. The prevalence of neck pain reported both for the previous 12 months and for the previous 7 d was 34% and more than half of them reported neck pain combined with pain in one or both arms. Analysis showed no significant association between neck pain and high WBV exposure; however, cases with neck pain more often experienced shocks and jolts in the vehicle as uncomfortable. There was no significant association between the 14 WBV measures and type of neck pain (neck pain vs. neck pain combined with arm pain). It seems as if characteristics of WBV exposure can explain neither existence nor the type of neck pain amongst professional drivers of forest machines. The logging industry is important for several industrialised countries. Drivers of forest machines frequently report neuromusculoskeletal pain from the neck. The type of neck pain is important for the decision of treatment modality and may be associated with exposure characteristics at work.

Computed tomography synthesis from magnetic resonance images in the pelvis using multiple random forests and auto-context features

NASA Astrophysics Data System (ADS)

Andreasen, Daniel; Edmund, Jens M.; Zografos, Vasileios; Menze, Bjoern H.; Van Leemput, Koen

2016-03-01

In radiotherapy treatment planning that is only based on magnetic resonance imaging (MRI), the electron density information usually obtained from computed tomography (CT) must be derived from the MRI by synthesizing a so-called pseudo CT (pCT). This is a non-trivial task since MRI intensities are neither uniquely nor quantitatively related to electron density. Typical approaches involve either a classification or regression model requiring specialized MRI sequences to solve intensity ambiguities, or an atlas-based model necessitating multiple registrations between atlases and subject scans. In this work, we explore a machine learning approach for creating a pCT of the pelvic region from conventional MRI sequences without using atlases. We use a random forest provided with information about local texture, edges and spatial features derived from the MRI. This helps to solve intensity ambiguities. Furthermore, we use the concept of auto-context by sequentially training a number of classification forests to create and improve context features, which are finally used to train a regression forest for pCT prediction. We evaluate the pCT quality in terms of the voxel-wise error and the radiologic accuracy as measured by water-equivalent path lengths. We compare the performance of our method against two baseline pCT strategies, which either set all MRI voxels in the subject equal to the CT value of water, or in addition transfer the bone volume from the real CT. We show an improved performance compared to both baseline pCTs suggesting that our method may be useful for MRI-only radiotherapy.

Reflections on the Development of a Machine Vision Technology for the Forest Products

Treesearch

Richard W. Conners; D.Earl Kline; Philip A. Araman; Robert L. Brisbon

1992-01-01

The authors have approximately 25 years experience in developing machine vision technology for the forest products industry. Based on this experience this paper will attempt to realistically predict what the future holds for this technology. In particular, this paper will attempt to describe some of the benefits this technology will offer, describe how the technology...

Producing landslide susceptibility maps by utilizing machine learning methods. The case of Finikas catchment basin, North Peloponnese, Greece.

NASA Astrophysics Data System (ADS)

Tsangaratos, Paraskevas; Ilia, Ioanna; Loupasakis, Constantinos; Papadakis, Michalis; Karimalis, Antonios

2017-04-01

The main objective of the present study was to apply two machine learning methods for the production of a landslide susceptibility map in the Finikas catchment basin, located in North Peloponnese, Greece and to compare their results. Specifically, Logistic Regression and Random Forest were utilized, based on a database of 40 sites classified into two categories, non-landslide and landslide areas that were separated into a training dataset (70% of the total data) and a validation dataset (remaining 30%). The identification of the areas was established by analyzing airborne imagery, extensive field investigation and the examination of previous research studies. Six landslide related variables were analyzed, namely: lithology, elevation, slope, aspect, distance to rivers and distance to faults. Within the Finikas catchment basin most of the reported landslides were located along the road network and within the residential complexes, classified as rotational and translational slides, and rockfalls, mainly caused due to the physical conditions and the general geotechnical behavior of the geological formation that cover the area. Each landslide susceptibility map was reclassified by applying the Geometric Interval classification technique into five classes, namely: very low susceptibility, low susceptibility, moderate susceptibility, high susceptibility, and very high susceptibility. The comparison and validation of the outcomes of each model were achieved using statistical evaluation measures, the receiving operating characteristic and the area under the success and predictive rate curves. The computation process was carried out using RStudio an integrated development environment for R language and ArcGIS 10.1 for compiling the data and producing the landslide susceptibility maps. From the outcomes of the Logistic Regression analysis it was induced that the highest b coefficient is allocated to lithology and slope, which was 2.8423 and 1.5841, respectively. From the estimation of the mean decrease in Gini coefficient performed during the application of Random Forest and the mean decrease in accuracy the most important variable is slope followed by lithology, aspect, elevation, distance from river network, and distance from faults, while the most used variables during the training phase were the variable aspect (21.45%), slope (20.53%) and lithology (19.84%). The outcomes of the analysis are consistent with previous studies concerning the area of research, which have indicated the high influence of lithology and slope in the manifestation of landslides. High percentage of landslide occurrence has been observed in Plio-Pleistocene sediments, flysch formations, and Cretaceous limestone. Also the presences of landslides have been associated with the degree of weathering and fragmentation, the orientation of the discontinuities surfaces and the intense morphological relief. The most accurate model was Random Forest which identified correctly 92.00% of the instances during the training phase, followed by the Logistic Regression 89.00%. The same pattern of accuracy was calculated during the validation phase, in which the Random Forest achieved a classification accuracy of 93.00%, while the Logistic Regression model achieved an accuracy of 91.00%. In conclusion, the outcomes of the study could be a useful cartographic product to local authorities and government agencies during the implementation of successful decision-making and land use planning strategies. Keywords: Landslide Susceptibility, Logistic Regression, Random Forest, GIS, Greece.

The Impact of Protein Structure and Sequence Similarity on the Accuracy of Machine-Learning Scoring Functions for Binding Affinity Prediction

PubMed Central

Peng, Jiangjun; Leung, Yee; Leung, Kwong-Sak; Wong, Man-Hon; Lu, Gang; Ballester, Pedro J.

2018-01-01

It has recently been claimed that the outstanding performance of machine-learning scoring functions (SFs) is exclusively due to the presence of training complexes with highly similar proteins to those in the test set. Here, we revisit this question using 24 similarity-based training sets, a widely used test set, and four SFs. Three of these SFs employ machine learning instead of the classical linear regression approach of the fourth SF (X-Score which has the best test set performance out of 16 classical SFs). We have found that random forest (RF)-based RF-Score-v3 outperforms X-Score even when 68% of the most similar proteins are removed from the training set. In addition, unlike X-Score, RF-Score-v3 is able to keep learning with an increasing training set size, becoming substantially more predictive than X-Score when the full 1105 complexes are used for training. These results show that machine-learning SFs owe a substantial part of their performance to training on complexes with dissimilar proteins to those in the test set, against what has been previously concluded using the same data. Given that a growing amount of structural and interaction data will be available from academic and industrial sources, this performance gap between machine-learning SFs and classical SFs is expected to enlarge in the future. PMID:29538331

Tear fluid proteomics multimarkers for diabetic retinopathy screening

PubMed Central

2013-01-01

Background The aim of the project was to develop a novel method for diabetic retinopathy screening based on the examination of tear fluid biomarker changes. In order to evaluate the usability of protein biomarkers for pre-screening purposes several different approaches were used, including machine learning algorithms. Methods All persons involved in the study had diabetes. Diabetic retinopathy (DR) was diagnosed by capturing 7-field fundus images, evaluated by two independent ophthalmologists. 165 eyes were examined (from 119 patients), 55 were diagnosed healthy and 110 images showed signs of DR. Tear samples were taken from all eyes and state-of-the-art nano-HPLC coupled ESI-MS/MS mass spectrometry protein identification was performed on all samples. Applicability of protein biomarkers was evaluated by six different optimally parameterized machine learning algorithms: Support Vector Machine, Recursive Partitioning, Random Forest, Naive Bayes, Logistic Regression, K-Nearest Neighbor. Results Out of the six investigated machine learning algorithms the result of Recursive Partitioning proved to be the most accurate. The performance of the system realizing the above algorithm reached 74% sensitivity and 48% specificity. Conclusions Protein biomarkers selected and classified with machine learning algorithms alone are at present not recommended for screening purposes because of low specificity and sensitivity values. This tool can be potentially used to improve the results of image processing methods as a complementary tool in automatic or semiautomatic systems. PMID:23919537

The Impact of Protein Structure and Sequence Similarity on the Accuracy of Machine-Learning Scoring Functions for Binding Affinity Prediction.

PubMed

Li, Hongjian; Peng, Jiangjun; Leung, Yee; Leung, Kwong-Sak; Wong, Man-Hon; Lu, Gang; Ballester, Pedro J

2018-03-14

It has recently been claimed that the outstanding performance of machine-learning scoring functions (SFs) is exclusively due to the presence of training complexes with highly similar proteins to those in the test set. Here, we revisit this question using 24 similarity-based training sets, a widely used test set, and four SFs. Three of these SFs employ machine learning instead of the classical linear regression approach of the fourth SF (X-Score which has the best test set performance out of 16 classical SFs). We have found that random forest (RF)-based RF-Score-v3 outperforms X-Score even when 68% of the most similar proteins are removed from the training set. In addition, unlike X-Score, RF-Score-v3 is able to keep learning with an increasing training set size, becoming substantially more predictive than X-Score when the full 1105 complexes are used for training. These results show that machine-learning SFs owe a substantial part of their performance to training on complexes with dissimilar proteins to those in the test set, against what has been previously concluded using the same data. Given that a growing amount of structural and interaction data will be available from academic and industrial sources, this performance gap between machine-learning SFs and classical SFs is expected to enlarge in the future.

Biodiversity mapping in a tropical West African forest with airborne hyperspectral data.

PubMed

Vaglio Laurin, Gaia; Cheung-Wai Chan, Jonathan; Chen, Qi; Lindsell, Jeremy A; Coomes, David A; Guerriero, Leila; Del Frate, Fabio; Miglietta, Franco; Valentini, Riccardo

2014-01-01

Tropical forests are major repositories of biodiversity, but are fast disappearing as land is converted to agriculture. Decision-makers need to know which of the remaining forests to prioritize for conservation, but the only spatial information on forest biodiversity has, until recently, come from a sparse network of ground-based plots. Here we explore whether airborne hyperspectral imagery can be used to predict the alpha diversity of upper canopy trees in a West African forest. The abundance of tree species were collected from 64 plots (each 1250 m(2) in size) within a Sierra Leonean national park, and Shannon-Wiener biodiversity indices were calculated. An airborne spectrometer measured reflectances of 186 bands in the visible and near-infrared spectral range at 1 m(2) resolution. The standard deviations of these reflectance values and their first-order derivatives were calculated for each plot from the c. 1250 pixels of hyperspectral information within them. Shannon-Wiener indices were then predicted from these plot-based reflectance statistics using a machine-learning algorithm (Random Forest). The regression model fitted the data well (pseudo-R(2) = 84.9%), and we show that standard deviations of green-band reflectances and infra-red region derivatives had the strongest explanatory powers. Our work shows that airborne hyperspectral sensing can be very effective at mapping canopy tree diversity, because its high spatial resolution allows within-plot heterogeneity in reflectance to be characterized, making it an effective tool for monitoring forest biodiversity over large geographic scales.

Biodiversity Mapping in a Tropical West African Forest with Airborne Hyperspectral Data

PubMed Central

Vaglio Laurin, Gaia; Chan, Jonathan Cheung-Wai; Chen, Qi; Lindsell, Jeremy A.; Coomes, David A.; Guerriero, Leila; Frate, Fabio Del; Miglietta, Franco; Valentini, Riccardo

2014-01-01

Tropical forests are major repositories of biodiversity, but are fast disappearing as land is converted to agriculture. Decision-makers need to know which of the remaining forests to prioritize for conservation, but the only spatial information on forest biodiversity has, until recently, come from a sparse network of ground-based plots. Here we explore whether airborne hyperspectral imagery can be used to predict the alpha diversity of upper canopy trees in a West African forest. The abundance of tree species were collected from 64 plots (each 1250 m2 in size) within a Sierra Leonean national park, and Shannon-Wiener biodiversity indices were calculated. An airborne spectrometer measured reflectances of 186 bands in the visible and near-infrared spectral range at 1 m2 resolution. The standard deviations of these reflectance values and their first-order derivatives were calculated for each plot from the c. 1250 pixels of hyperspectral information within them. Shannon-Wiener indices were then predicted from these plot-based reflectance statistics using a machine-learning algorithm (Random Forest). The regression model fitted the data well (pseudo-R2 = 84.9%), and we show that standard deviations of green-band reflectances and infra-red region derivatives had the strongest explanatory powers. Our work shows that airborne hyperspectral sensing can be very effective at mapping canopy tree diversity, because its high spatial resolution allows within-plot heterogeneity in reflectance to be characterized, making it an effective tool for monitoring forest biodiversity over large geographic scales. PMID:24937407

A machine learning-based framework to identify type 2 diabetes through electronic health records

PubMed Central

Zheng, Tao; Xie, Wei; Xu, Liling; He, Xiaoying; Zhang, Ya; You, Mingrong; Yang, Gong; Chen, You

2016-01-01

Objective To discover diverse genotype-phenotype associations affiliated with Type 2 Diabetes Mellitus (T2DM) via genome-wide association study (GWAS) and phenome-wide association study (PheWAS), more cases (T2DM subjects) and controls (subjects without T2DM) are required to be identified (e.g., via Electronic Health Records (EHR)). However, existing expert based identification algorithms often suffer in a low recall rate and could miss a large number of valuable samples under conservative filtering standards. The goal of this work is to develop a semi-automated framework based on machine learning as a pilot study to liberalize filtering criteria to improve recall rate with a keeping of low false positive rate. Materials and methods We propose a data informed framework for identifying subjects with and without T2DM from EHR via feature engineering and machine learning. We evaluate and contrast the identification performance of widely-used machine learning models within our framework, including k-Nearest-Neighbors, Naïve Bayes, Decision Tree, Random Forest, Support Vector Machine and Logistic Regression. Our framework was conducted on 300 patient samples (161 cases, 60 controls and 79 unconfirmed subjects), randomly selected from 23,281 diabetes related cohort retrieved from a regional distributed EHR repository ranging from 2012 to 2014. Results We apply top-performing machine learning algorithms on the engineered features. We benchmark and contrast the accuracy, precision, AUC, sensitivity and specificity of classification models against the state-of-the-art expert algorithm for identification of T2DM subjects. Our results indicate that the framework achieved high identification performances (∼0.98 in average AUC), which are much higher than the state-of-the-art algorithm (0.71 in AUC). Discussion Expert algorithm-based identification of T2DM subjects from EHR is often hampered by the high missing rates due to their conservative selection criteria. Our framework leverages machine learning and feature engineering to loosen such selection criteria to achieve a high identification rate of cases and controls. Conclusions Our proposed framework demonstrates a more accurate and efficient approach for identifying subjects with and without T2DM from EHR. PMID:27919371

A machine learning-based framework to identify type 2 diabetes through electronic health records.

PubMed

Zheng, Tao; Xie, Wei; Xu, Liling; He, Xiaoying; Zhang, Ya; You, Mingrong; Yang, Gong; Chen, You

2017-01-01

To discover diverse genotype-phenotype associations affiliated with Type 2 Diabetes Mellitus (T2DM) via genome-wide association study (GWAS) and phenome-wide association study (PheWAS), more cases (T2DM subjects) and controls (subjects without T2DM) are required to be identified (e.g., via Electronic Health Records (EHR)). However, existing expert based identification algorithms often suffer in a low recall rate and could miss a large number of valuable samples under conservative filtering standards. The goal of this work is to develop a semi-automated framework based on machine learning as a pilot study to liberalize filtering criteria to improve recall rate with a keeping of low false positive rate. We propose a data informed framework for identifying subjects with and without T2DM from EHR via feature engineering and machine learning. We evaluate and contrast the identification performance of widely-used machine learning models within our framework, including k-Nearest-Neighbors, Naïve Bayes, Decision Tree, Random Forest, Support Vector Machine and Logistic Regression. Our framework was conducted on 300 patient samples (161 cases, 60 controls and 79 unconfirmed subjects), randomly selected from 23,281 diabetes related cohort retrieved from a regional distributed EHR repository ranging from 2012 to 2014. We apply top-performing machine learning algorithms on the engineered features. We benchmark and contrast the accuracy, precision, AUC, sensitivity and specificity of classification models against the state-of-the-art expert algorithm for identification of T2DM subjects. Our results indicate that the framework achieved high identification performances (∼0.98 in average AUC), which are much higher than the state-of-the-art algorithm (0.71 in AUC). Expert algorithm-based identification of T2DM subjects from EHR is often hampered by the high missing rates due to their conservative selection criteria. Our framework leverages machine learning and feature engineering to loosen such selection criteria to achieve a high identification rate of cases and controls. Our proposed framework demonstrates a more accurate and efficient approach for identifying subjects with and without T2DM from EHR. Copyright © 2016 Elsevier Ireland Ltd. All rights reserved.

Unsupervised learning on scientific ocean drilling datasets from the South China Sea

NASA Astrophysics Data System (ADS)

Tse, Kevin C.; Chiu, Hon-Chim; Tsang, Man-Yin; Li, Yiliang; Lam, Edmund Y.

2018-06-01

Unsupervised learning methods were applied to explore data patterns in multivariate geophysical datasets collected from ocean floor sediment core samples coming from scientific ocean drilling in the South China Sea. Compared to studies on similar datasets, but using supervised learning methods which are designed to make predictions based on sample training data, unsupervised learning methods require no a priori information and focus only on the input data. In this study, popular unsupervised learning methods including K-means, self-organizing maps, hierarchical clustering and random forest were coupled with different distance metrics to form exploratory data clusters. The resulting data clusters were externally validated with lithologic units and geologic time scales assigned to the datasets by conventional methods. Compact and connected data clusters displayed varying degrees of correspondence with existing classification by lithologic units and geologic time scales. K-means and self-organizing maps were observed to perform better with lithologic units while random forest corresponded best with geologic time scales. This study sets a pioneering example of how unsupervised machine learning methods can be used as an automatic processing tool for the increasingly high volume of scientific ocean drilling data.

Detection of compensatory balance responses using wearable electromyography sensors for fall-risk assessment.

PubMed

Nouredanesh, Mina; Kukreja, Sunil L; Tung, James

2016-08-01

Loss of balance is prevalent in older adults and populations with gait and balance impairments. The present paper aims to develop a method to automatically distinguish compensatory balance responses (CBRs) from normal gait, based on activity patterns of muscles involved in maintaining balance. In this study, subjects were perturbed by lateral pushes while walking and surface electromyography (sEMG) signals were recorded from four muscles in their right leg. To extract sEMG time domain features, several filtering characteristics and segmentation approaches are examined. The performance of three classification methods, i.e., k-nearest neighbor, support vector machines, and random forests, were investigated for accurate detection of CBRs. Our results show that features extracted in the 50-200Hz band, segmented using peak sEMG amplitudes, and a random forest classifier detected CBRs with an accuracy of 92.35%. Moreover, our results support the important role of biceps femoris and rectus femoris muscles in stabilization and consequently discerning CBRs. This study contributes towards the development of wearable sensor systems to accurately and reliably monitor gait and balance control behavior in at-home settings (unsupervised conditions), over long periods of time, towards personalized fall risk assessment tools.

Estimation of sleep status in sleep apnea patients using a novel head actigraphy technique.

PubMed

Hummel, Richard; Bradley, T Douglas; Fernie, Geoff R; Chang, S J Isaac; Alshaer, Hisham

2015-01-01

Polysomnography is a comprehensive modality for diagnosing sleep apnea (SA), but it is expensive and not widely available. Several technologies have been developed for portable diagnosis of SA in the home, most of which lack the ability to detect sleep status. Wrist actigraphy (accelerometry) has been adopted to cover this limitation. However, head actigraphy has not been systematically evaluated for this purpose. Therefore, the aim of this study was to evaluate the ability of head actigraphy to detect sleep/wake status. We obtained full overnight 3-axis head accelerometry data from 75 sleep apnea patient recordings. These were split into training and validation groups (2:1). Data were preprocessed and 5 features were extracted. Different feature combinations were fed into 3 different classifiers, namely support vector machine, logistic regression, and random forests, each of which was trained and validated on a different subgroup. The random forest algorithm yielded the highest performance, with an area under the receiver operating characteristic (ROC) curve of 0.81 for detection of sleep status. This shows that this technique has a very good performance in detecting sleep status in SA patients despite the specificities in this population, such as respiration related movements.

Comparison of classification methods for voxel-based prediction of acute ischemic stroke outcome following intra-arterial intervention

NASA Astrophysics Data System (ADS)

Winder, Anthony J.; Siemonsen, Susanne; Flottmann, Fabian; Fiehler, Jens; Forkert, Nils D.

2017-03-01

Voxel-based tissue outcome prediction in acute ischemic stroke patients is highly relevant for both clinical routine and research. Previous research has shown that features extracted from baseline multi-parametric MRI datasets have a high predictive value and can be used for the training of classifiers, which can generate tissue outcome predictions for both intravenous and conservative treatments. However, with the recent advent and popularization of intra-arterial thrombectomy treatment, novel research specifically addressing the utility of predictive classi- fiers for thrombectomy intervention is necessary for a holistic understanding of current stroke treatment options. The aim of this work was to develop three clinically viable tissue outcome prediction models using approximate nearest-neighbor, generalized linear model, and random decision forest approaches and to evaluate the accuracy of predicting tissue outcome after intra-arterial treatment. Therefore, the three machine learning models were trained, evaluated, and compared using datasets of 42 acute ischemic stroke patients treated with intra-arterial thrombectomy. Classifier training utilized eight voxel-based features extracted from baseline MRI datasets and five global features. Evaluation of classifier-based predictions was performed via comparison to the known tissue outcome, which was determined in follow-up imaging, using the Dice coefficient and leave-on-patient-out cross validation. The random decision forest prediction model led to the best tissue outcome predictions with a mean Dice coefficient of 0.37. The approximate nearest-neighbor and generalized linear model performed equally suboptimally with average Dice coefficients of 0.28 and 0.27 respectively, suggesting that both non-linearity and machine learning are desirable properties of a classifier well-suited to the intra-arterial tissue outcome prediction problem.

A domain analysis approach to clear-air turbulence forecasting using high-density in-situ measurements

NASA Astrophysics Data System (ADS)

Abernethy, Jennifer A.

Pilots' ability to avoid clear-air turbulence (CAT) during flight affects the safety of the millions of people who fly commercial airlines and other aircraft, and turbulence costs millions in injuries and aircraft maintenance every year. Forecasting CAT is not straightforward, however; microscale features like the turbulence eddies that affect aircraft (100m) are below the current resolution of operational numerical weather prediction (NWP) models, and the only evidence of CAT episodes, until recently, has been sparse, subjective reports from pilots known as PIREPs. To forecast CAT, researchers use a simple weighted sum of top-performing turbulence indicators derived from NWP model outputs---termed diagnostics---based on their agreement with current PIREPs. However, a new, quantitative source of observation data---high-density measurements made by sensor equipment and software on aircraft, called in-situ measurements---is now available. The main goal of this thesis is to develop new data analysis and processing techniques to apply to the model and new observation data, in order to improve CAT forecasting accuracy. This thesis shows that using in-situ data improves forecasting accuracy and that automated machine learning algorithms such as support vector machines (SVM), logistic regression, and random forests, can match current performance while eliminating almost all hand-tuning. Feature subset selection is paired with the new algorithms to choose diagnostics that predict well as a group rather than individually. Specializing forecasts and choice of diagnostics by geographic region further improves accuracy because of the geographic variation in turbulence sources. This work uses random forests to find climatologically-relevant regions based on these variations and implements a forecasting system testbed which brings these techniques together to rapidly prototype new, regionalized versions of operational CAT forecasting systems.

Integrated Application of Random Forest and Artificial Neural Network Algorithms to Predict Viral Contamination in Coastal Waters

NASA Astrophysics Data System (ADS)

Shamkhali Chenar, S.; Deng, Z.

2017-12-01

Pathogenic viruses pose a significant public health threat and economic losses to shellfish industry in the coastal environment. Norovirus is a contagious virus and the leading cause of epidemic gastroenteritis following consumption of oysters harvested from sewage-contaminated waters. While it is challenging to detect noroviruses in coastal waters due to the lack of sensitive and routine diagnostic methods, machine learning techniques are allowing us to prevent or at least reduce the risks by developing effective predictive models. This study attempts to develop an algorithm between historical norovirus outbreak reports and environmental parameters including water temperature, solar radiation, water level, salinity, precipitation, and wind. For this purpose, the Random Forests statistical technique was utilized to select relevant environmental parameters and their various combinations with different time lags controlling the virus distribution in oyster harvesting areas along the Louisiana Coast. An Artificial Neural Networks (ANN) approach was then presented to predict the outbreaks using a final set of input variables. Finally, a sensitivity analysis was conducted to evaluate relative importance and contribution of the input variables to the model output. Findings demonstrated that the developed model was capable of reproducing historical oyster norovirus outbreaks along the Louisiana Coast with the overall accuracy of than 99.83%, demonstrating the efficacy of the model. Moreover, the increase in water temperature, solar radiation, water level, and salinity, and the decrease in wind and rainfall are associated with the reduction in the model-predicted risk of norovirus outbreak according to sensitivity analysis results. In conclusion, the presented machine learning approach provided reliable tools for predicting potential norovirus outbreaks and could be used for early detection of possible outbreaks and reduce the risk of norovirus to public health and the seafood industry.

AutoIHC-scoring: a machine learning framework for automated Allred scoring of molecular expression in ER- and PR-stained breast cancer tissue.

PubMed

Tewary, S; Arun, I; Ahmed, R; Chatterjee, S; Chakraborty, C

2017-11-01

In prognostic evaluation of breast cancer Immunohistochemical (IHC) markers namely, oestrogen receptor (ER) and progesterone receptor (PR) are widely used. The expert pathologist investigates qualitatively the stained tissue slide under microscope to provide the Allred score; which is clinically used for therapeutic decision making. Such qualitative judgment is time-consuming, tedious and more often suffers from interobserver variability. As a result, it leads to imprecise IHC score for ER and PR. To overcome this, there is an urgent need of developing a reliable and efficient IHC quantifier for high throughput decision making. In view of this, our study aims at developing an automated IHC profiler for quantitative assessment of ER and PR molecular expression from stained tissue images. We propose here to use CMYK colour space for positively and negatively stained cell extraction for proportion score. Also colour features are used for quantitative assessment of intensity scoring among the positively stained cells. Five different machine learning models namely artificial neural network, Naïve Bayes, K-nearest neighbours, decision tree and random forest are considered for learning the colour features using average red, green and blue pixel values of positively stained cell patches. Fifty cases of ER- and PR-stained tissues have been evaluated for validation with the expert pathologist's score. All five models perform adequately where random forest shows the best correlation with the expert's score (Pearson's correlation coefficient = 0.9192). In the proposed approach the average variation of diaminobenzidine (DAB) to nuclear area from the expert's score is found to be 7.58%, as compared to 27.83% for state-of-the-art ImmunoRatio software. © 2017 The Authors Journal of Microscopy © 2017 Royal Microscopical Society.

Empirical study of seven data mining algorithms on different characteristics of datasets for biomedical classification applications.

PubMed

Zhang, Yiyan; Xin, Yi; Li, Qin; Ma, Jianshe; Li, Shuai; Lv, Xiaodan; Lv, Weiqi

2017-11-02

Various kinds of data mining algorithms are continuously raised with the development of related disciplines. The applicable scopes and their performances of these algorithms are different. Hence, finding a suitable algorithm for a dataset is becoming an important emphasis for biomedical researchers to solve practical problems promptly. In this paper, seven kinds of sophisticated active algorithms, namely, C4.5, support vector machine, AdaBoost, k-nearest neighbor, naïve Bayes, random forest, and logistic regression, were selected as the research objects. The seven algorithms were applied to the 12 top-click UCI public datasets with the task of classification, and their performances were compared through induction and analysis. The sample size, number of attributes, number of missing values, and the sample size of each class, correlation coefficients between variables, class entropy of task variable, and the ratio of the sample size of the largest class to the least class were calculated to character the 12 research datasets. The two ensemble algorithms reach high accuracy of classification on most datasets. Moreover, random forest performs better than AdaBoost on the unbalanced dataset of the multi-class task. Simple algorithms, such as the naïve Bayes and logistic regression model are suitable for a small dataset with high correlation between the task and other non-task attribute variables. K-nearest neighbor and C4.5 decision tree algorithms perform well on binary- and multi-class task datasets. Support vector machine is more adept on the balanced small dataset of the binary-class task. No algorithm can maintain the best performance in all datasets. The applicability of the seven data mining algorithms on the datasets with different characteristics was summarized to provide a reference for biomedical researchers or beginners in different fields.

Resolving Transition Metal Chemical Space: Feature Selection for Machine Learning and Structure-Property Relationships.

PubMed

Janet, Jon Paul; Kulik, Heather J

2017-11-22

Machine learning (ML) of quantum mechanical properties shows promise for accelerating chemical discovery. For transition metal chemistry where accurate calculations are computationally costly and available training data sets are small, the molecular representation becomes a critical ingredient in ML model predictive accuracy. We introduce a series of revised autocorrelation functions (RACs) that encode relationships of the heuristic atomic properties (e.g., size, connectivity, and electronegativity) on a molecular graph. We alter the starting point, scope, and nature of the quantities evaluated in standard ACs to make these RACs amenable to inorganic chemistry. On an organic molecule set, we first demonstrate superior standard AC performance to other presently available topological descriptors for ML model training, with mean unsigned errors (MUEs) for atomization energies on set-aside test molecules as low as 6 kcal/mol. For inorganic chemistry, our RACs yield 1 kcal/mol ML MUEs on set-aside test molecules in spin-state splitting in comparison to 15-20× higher errors for feature sets that encode whole-molecule structural information. Systematic feature selection methods including univariate filtering, recursive feature elimination, and direct optimization (e.g., random forest and LASSO) are compared. Random-forest- or LASSO-selected subsets 4-5× smaller than the full RAC set produce sub- to 1 kcal/mol spin-splitting MUEs, with good transferability to metal-ligand bond length prediction (0.004-5 Å MUE) and redox potential on a smaller data set (0.2-0.3 eV MUE). Evaluation of feature selection results across property sets reveals the relative importance of local, electronic descriptors (e.g., electronegativity, atomic number) in spin-splitting and distal, steric effects in redox potential and bond lengths.

Characterization of cervigram image sharpness using multiple self-referenced measurements and random forest classifiers

NASA Astrophysics Data System (ADS)

Jaiswal, Mayoore; Horning, Matt; Hu, Liming; Ben-Or, Yau; Champlin, Cary; Wilson, Benjamin; Levitz, David

2018-02-01

Cervical cancer is the fourth most common cancer among women worldwide and is especially prevalent in low resource settings due to lack of screening and treatment options. Visual inspection with acetic acid (VIA) is a widespread and cost-effective screening method for cervical pre-cancer lesions, but accuracy depends on the experience level of the health worker. Digital cervicography, capturing images of the cervix, enables review by an off-site expert or potentially a machine learning algorithm. These reviews require images of sufficient quality. However, image quality varies greatly across users. A novel algorithm was developed to evaluate the sharpness of images captured with the MobileODT's digital cervicography device (EVA System), in order to, eventually provide feedback to the health worker. The key challenges are that the algorithm evaluates only a single image of each cervix, it needs to be robust to the variability in cervix images and fast enough to run in real time on a mobile device, and the machine learning model needs to be small enough to fit on a mobile device's memory, train on a small imbalanced dataset and run in real-time. In this paper, the focus scores of a preprocessed image and a Gaussian-blurred version of the image are calculated using established methods and used as features. A feature selection metric is proposed to select the top features which were then used in a random forest classifier to produce the final focus score. The resulting model, based on nine calculated focus scores, achieved significantly better accuracy than any single focus measure when tested on a holdout set of images. The area under the receiver operating characteristics curve was 0.9459.

A study of the effectiveness of machine learning methods for classification of clinical interview fragments into a large number of categories.

PubMed

Hasan, Mehedi; Kotov, Alexander; Carcone, April; Dong, Ming; Naar, Sylvie; Hartlieb, Kathryn Brogan

2016-08-01

This study examines the effectiveness of state-of-the-art supervised machine learning methods in conjunction with different feature types for the task of automatic annotation of fragments of clinical text based on codebooks with a large number of categories. We used a collection of motivational interview transcripts consisting of 11,353 utterances, which were manually annotated by two human coders as the gold standard, and experimented with state-of-art classifiers, including Naïve Bayes, J48 Decision Tree, Support Vector Machine (SVM), Random Forest (RF), AdaBoost, DiscLDA, Conditional Random Fields (CRF) and Convolutional Neural Network (CNN) in conjunction with lexical, contextual (label of the previous utterance) and semantic (distribution of words in the utterance across the Linguistic Inquiry and Word Count dictionaries) features. We found out that, when the number of classes is large, the performance of CNN and CRF is inferior to SVM. When only lexical features were used, interview transcripts were automatically annotated by SVM with the highest classification accuracy among all classifiers of 70.8%, 61% and 53.7% based on the codebooks consisting of 17, 20 and 41 codes, respectively. Using contextual and semantic features, as well as their combination, in addition to lexical ones, improved the accuracy of SVM for annotation of utterances in motivational interview transcripts with a codebook consisting of 17 classes to 71.5%, 74.2%, and 75.1%, respectively. Our results demonstrate the potential of using machine learning methods in conjunction with lexical, semantic and contextual features for automatic annotation of clinical interview transcripts with near-human accuracy. Copyright © 2016 Elsevier Inc. All rights reserved.

Discrimination of plant root zone water status in greenhouse production based on phenotyping and machine learning techniques.

PubMed

Guo, Doudou; Juan, Jiaxiang; Chang, Liying; Zhang, Jingjin; Huang, Danfeng

2017-08-15

Plant-based sensing on water stress can provide sensitive and direct reference for precision irrigation system in greenhouse. However, plant information acquisition, interpretation, and systematical application remain insufficient. This study developed a discrimination method for plant root zone water status in greenhouse by integrating phenotyping and machine learning techniques. Pakchoi plants were used and treated by three root zone moisture levels, 40%, 60%, and 80% relative water content. Three classification models, Random Forest (RF), Neural Network (NN), and Support Vector Machine (SVM) were developed and validated in different scenarios with overall accuracy over 90% for all. SVM model had the highest value, but it required the longest training time. All models had accuracy over 85% in all scenarios, and more stable performance was observed in RF model. Simplified SVM model developed by the top five most contributing traits had the largest accuracy reduction as 29.5%, while simplified RF and NN model still maintained approximately 80%. For real case application, factors such as operation cost, precision requirement, and system reaction time should be synthetically considered in model selection. Our work shows it is promising to discriminate plant root zone water status by implementing phenotyping and machine learning techniques for precision irrigation management.

Machine Learning methods for Quantitative Radiomic Biomarkers.

PubMed

Parmar, Chintan; Grossmann, Patrick; Bussink, Johan; Lambin, Philippe; Aerts, Hugo J W L

2015-08-17

Radiomics extracts and mines large number of medical imaging features quantifying tumor phenotypic characteristics. Highly accurate and reliable machine-learning approaches can drive the success of radiomic applications in clinical care. In this radiomic study, fourteen feature selection methods and twelve classification methods were examined in terms of their performance and stability for predicting overall survival. A total of 440 radiomic features were extracted from pre-treatment computed tomography (CT) images of 464 lung cancer patients. To ensure the unbiased evaluation of different machine-learning methods, publicly available implementations along with reported parameter configurations were used. Furthermore, we used two independent radiomic cohorts for training (n = 310 patients) and validation (n = 154 patients). We identified that Wilcoxon test based feature selection method WLCX (stability = 0.84 ± 0.05, AUC = 0.65 ± 0.02) and a classification method random forest RF (RSD = 3.52%, AUC = 0.66 ± 0.03) had highest prognostic performance with high stability against data perturbation. Our variability analysis indicated that the choice of classification method is the most dominant source of performance variation (34.21% of total variance). Identification of optimal machine-learning methods for radiomic applications is a crucial step towards stable and clinically relevant radiomic biomarkers, providing a non-invasive way of quantifying and monitoring tumor-phenotypic characteristics in clinical practice.

Detection of Periodic Leg Movements by Machine Learning Methods Using Polysomnographic Parameters Other Than Leg Electromyography

PubMed Central

Umut, İlhan; Çentik, Güven

2016-01-01

The number of channels used for polysomnographic recording frequently causes difficulties for patients because of the many cables connected. Also, it increases the risk of having troubles during recording process and increases the storage volume. In this study, it is intended to detect periodic leg movement (PLM) in sleep with the use of the channels except leg electromyography (EMG) by analysing polysomnography (PSG) data with digital signal processing (DSP) and machine learning methods. PSG records of 153 patients of different ages and genders with PLM disorder diagnosis were examined retrospectively. A novel software was developed for the analysis of PSG records. The software utilizes the machine learning algorithms, statistical methods, and DSP methods. In order to classify PLM, popular machine learning methods (multilayer perceptron, K-nearest neighbour, and random forests) and logistic regression were used. Comparison of classified results showed that while K-nearest neighbour classification algorithm had higher average classification rate (91.87%) and lower average classification error value (RMSE = 0.2850), multilayer perceptron algorithm had the lowest average classification rate (83.29%) and the highest average classification error value (RMSE = 0.3705). Results showed that PLM can be classified with high accuracy (91.87%) without leg EMG record being present. PMID:27213008

Detection of Periodic Leg Movements by Machine Learning Methods Using Polysomnographic Parameters Other Than Leg Electromyography.

PubMed

Umut, İlhan; Çentik, Güven

2016-01-01

The number of channels used for polysomnographic recording frequently causes difficulties for patients because of the many cables connected. Also, it increases the risk of having troubles during recording process and increases the storage volume. In this study, it is intended to detect periodic leg movement (PLM) in sleep with the use of the channels except leg electromyography (EMG) by analysing polysomnography (PSG) data with digital signal processing (DSP) and machine learning methods. PSG records of 153 patients of different ages and genders with PLM disorder diagnosis were examined retrospectively. A novel software was developed for the analysis of PSG records. The software utilizes the machine learning algorithms, statistical methods, and DSP methods. In order to classify PLM, popular machine learning methods (multilayer perceptron, K-nearest neighbour, and random forests) and logistic regression were used. Comparison of classified results showed that while K-nearest neighbour classification algorithm had higher average classification rate (91.87%) and lower average classification error value (RMSE = 0.2850), multilayer perceptron algorithm had the lowest average classification rate (83.29%) and the highest average classification error value (RMSE = 0.3705). Results showed that PLM can be classified with high accuracy (91.87%) without leg EMG record being present.

Classification of Strawberry Fruit Shape by Machine Learning

NASA Astrophysics Data System (ADS)

Ishikawa, T.; Hayashi, A.; Nagamatsu, S.; Kyutoku, Y.; Dan, I.; Wada, T.; Oku, K.; Saeki, Y.; Uto, T.; Tanabata, T.; Isobe, S.; Kochi, N.

2018-05-01

Shape is one of the most important traits of agricultural products due to its relationships with the quality, quantity, and value of the products. For strawberries, the nine types of fruit shape were defined and classified by humans based on the sampler patterns of the nine types. In this study, we tested the classification of strawberry shapes by machine learning in order to increase the accuracy of the classification, and we introduce the concept of computerization into this field. Four types of descriptors were extracted from the digital images of strawberries: (1) the Measured Values (MVs) including the length of the contour line, the area, the fruit length and width, and the fruit width/length ratio; (2) the Ellipse Similarity Index (ESI); (3) Elliptic Fourier Descriptors (EFDs), and (4) Chain Code Subtraction (CCS). We used these descriptors for the classification test along with the random forest approach, and eight of the nine shape types were classified with combinations of MVs + CCS + EFDs. CCS is a descriptor that adds human knowledge to the chain codes, and it showed higher robustness in classification than the other descriptors. Our results suggest machine learning's high ability to classify fruit shapes accurately. We will attempt to increase the classification accuracy and apply the machine learning methods to other plant species.

Text mining approach to predict hospital admissions using early medical records from the emergency department.

PubMed

Lucini, Filipe R; S Fogliatto, Flavio; C da Silveira, Giovani J; L Neyeloff, Jeruza; Anzanello, Michel J; de S Kuchenbecker, Ricardo; D Schaan, Beatriz

2017-04-01

Emergency department (ED) overcrowding is a serious issue for hospitals. Early information on short-term inward bed demand from patients receiving care at the ED may reduce the overcrowding problem, and optimize the use of hospital resources. In this study, we use text mining methods to process data from early ED patient records using the SOAP framework, and predict future hospitalizations and discharges. We try different approaches for pre-processing of text records and to predict hospitalization. Sets-of-words are obtained via binary representation, term frequency, and term frequency-inverse document frequency. Unigrams, bigrams and trigrams are tested for feature formation. Feature selection is based on χ 2 and F-score metrics. In the prediction module, eight text mining methods are tested: Decision Tree, Random Forest, Extremely Randomized Tree, AdaBoost, Logistic Regression, Multinomial Naïve Bayes, Support Vector Machine (Kernel linear) and Nu-Support Vector Machine (Kernel linear). Prediction performance is evaluated by F1-scores. Precision and Recall values are also informed for all text mining methods tested. Nu-Support Vector Machine was the text mining method with the best overall performance. Its average F1-score in predicting hospitalization was 77.70%, with a standard deviation (SD) of 0.66%. The method could be used to manage daily routines in EDs such as capacity planning and resource allocation. Text mining could provide valuable information and facilitate decision-making by inward bed management teams. Copyright © 2017 Elsevier Ireland Ltd. All rights reserved.

Machine learning methods reveal the temporal pattern of dengue incidence using meteorological factors in metropolitan Manila, Philippines.

PubMed

Carvajal, Thaddeus M; Viacrusis, Katherine M; Hernandez, Lara Fides T; Ho, Howell T; Amalin, Divina M; Watanabe, Kozo

2018-04-17

Several studies have applied ecological factors such as meteorological variables to develop models and accurately predict the temporal pattern of dengue incidence or occurrence. With the vast amount of studies that investigated this premise, the modeling approaches differ from each study and only use a single statistical technique. It raises the question of whether which technique would be robust and reliable. Hence, our study aims to compare the predictive accuracy of the temporal pattern of Dengue incidence in Metropolitan Manila as influenced by meteorological factors from four modeling techniques, (a) General Additive Modeling, (b) Seasonal Autoregressive Integrated Moving Average with exogenous variables (c) Random Forest and (d) Gradient Boosting. Dengue incidence and meteorological data (flood, precipitation, temperature, southern oscillation index, relative humidity, wind speed and direction) of Metropolitan Manila from January 1, 2009 - December 31, 2013 were obtained from respective government agencies. Two types of datasets were used in the analysis; observed meteorological factors (MF) and its corresponding delayed or lagged effect (LG). After which, these datasets were subjected to the four modeling techniques. The predictive accuracy and variable importance of each modeling technique were calculated and evaluated. Among the statistical modeling techniques, Random Forest showed the best predictive accuracy. Moreover, the delayed or lag effects of the meteorological variables was shown to be the best dataset to use for such purpose. Thus, the model of Random Forest with delayed meteorological effects (RF-LG) was deemed the best among all assessed models. Relative humidity was shown to be the top-most important meteorological factor in the best model. The study exhibited that there are indeed different predictive outcomes generated from each statistical modeling technique and it further revealed that the Random forest model with delayed meteorological effects to be the best in predicting the temporal pattern of Dengue incidence in Metropolitan Manila. It is also noteworthy that the study also identified relative humidity as an important meteorological factor along with rainfall and temperature that can influence this temporal pattern.

Quantifying Biomass from Point Clouds by Connecting Representations of Ecosystem Structure

NASA Astrophysics Data System (ADS)

Hendryx, S. M.; Barron-Gafford, G.

2017-12-01

Quantifying terrestrial ecosystem biomass is an essential part of monitoring carbon stocks and fluxes within the global carbon cycle and optimizing natural resource management. Point cloud data such as from lidar and structure from motion can be effective for quantifying biomass over large areas, but significant challenges remain in developing effective models that allow for such predictions. Inference models that estimate biomass from point clouds are established in many environments, yet, are often scale-dependent, needing to be fitted and applied at the same spatial scale and grid size at which they were developed. Furthermore, training such models typically requires large in situ datasets that are often prohibitively costly or time-consuming to obtain. We present here a scale- and sensor-invariant framework for efficiently estimating biomass from point clouds. Central to this framework, we present a new algorithm, assignPointsToExistingClusters, that has been developed for finding matches between in situ data and clusters in remotely-sensed point clouds. The algorithm can be used for assessing canopy segmentation accuracy and for training and validating machine learning models for predicting biophysical variables. We demonstrate the algorithm's efficacy by using it to train a random forest model of above ground biomass in a shrubland environment in Southern Arizona. We show that by learning a nonlinear function to estimate biomass from segmented canopy features we can reduce error, especially in the presence of inaccurate clusterings, when compared to a traditional, deterministic technique to estimate biomass from remotely measured canopies. Our random forest on cluster features model extends established methods of training random forest regressions to predict biomass of subplots but requires significantly less training data and is scale invariant. The random forest on cluster features model reduced mean absolute error, when evaluated on all test data in leave one out cross validation, by 40.6% from deterministic mesquite allometry and 35.9% from the inferred ecosystem-state allometric function. Our framework should allow for the inference of biomass more efficiently than common subplot methods and more accurately than individual tree segmentation methods in densely vegetated environments.

Calculating utilization rates for rubber tired grapple skidders in the Southern United States

Treesearch

Jason D. Thompson

2001-01-01

Utilization rate is an important factor in calculating machine rates for forest harvesting machines. Machine rates allow an evaluation of harvesting system costs and facilitate comparisons between different systems and machines. There are many factors that affect utilization rate. These include mechanical delays, non-mechanical delays, operational lost time, and...

A Hybrid Color Space for Skin Detection Using Genetic Algorithm Heuristic Search and Principal Component Analysis Technique

PubMed Central

2015-01-01

Color is one of the most prominent features of an image and used in many skin and face detection applications. Color space transformation is widely used by researchers to improve face and skin detection performance. Despite the substantial research efforts in this area, choosing a proper color space in terms of skin and face classification performance which can address issues like illumination variations, various camera characteristics and diversity in skin color tones has remained an open issue. This research proposes a new three-dimensional hybrid color space termed SKN by employing the Genetic Algorithm heuristic and Principal Component Analysis to find the optimal representation of human skin color in over seventeen existing color spaces. Genetic Algorithm heuristic is used to find the optimal color component combination setup in terms of skin detection accuracy while the Principal Component Analysis projects the optimal Genetic Algorithm solution to a less complex dimension. Pixel wise skin detection was used to evaluate the performance of the proposed color space. We have employed four classifiers including Random Forest, Naïve Bayes, Support Vector Machine and Multilayer Perceptron in order to generate the human skin color predictive model. The proposed color space was compared to some existing color spaces and shows superior results in terms of pixel-wise skin detection accuracy. Experimental results show that by using Random Forest classifier, the proposed SKN color space obtained an average F-score and True Positive Rate of 0.953 and False Positive Rate of 0.0482 which outperformed the existing color spaces in terms of pixel wise skin detection accuracy. The results also indicate that among the classifiers used in this study, Random Forest is the most suitable classifier for pixel wise skin detection applications. PMID:26267377

Mediastinal lymph node detection and station mapping on chest CT using spatial priors and random forest

DOE Office of Scientific and Technical Information (OSTI.GOV)

Liu, Jiamin; Hoffman, Joanne; Zhao, Jocelyn

2016-07-15

Purpose: To develop an automated system for mediastinal lymph node detection and station mapping for chest CT. Methods: The contextual organs, trachea, lungs, and spine are first automatically identified to locate the region of interest (ROI) (mediastinum). The authors employ shape features derived from Hessian analysis, local object scale, and circular transformation that are computed per voxel in the ROI. Eight more anatomical structures are simultaneously segmented by multiatlas label fusion. Spatial priors are defined as the relative multidimensional distance vectors corresponding to each structure. Intensity, shape, and spatial prior features are integrated and parsed by a random forest classifiermore » for lymph node detection. The detected candidates are then segmented by the following curve evolution process. Texture features are computed on the segmented lymph nodes and a support vector machine committee is used for final classification. For lymph node station labeling, based on the segmentation results of the above anatomical structures, the textual definitions of mediastinal lymph node map according to the International Association for the Study of Lung Cancer are converted into patient-specific color-coded CT image, where the lymph node station can be automatically assigned for each detected node. Results: The chest CT volumes from 70 patients with 316 enlarged mediastinal lymph nodes are used for validation. For lymph node detection, their system achieves 88% sensitivity at eight false positives per patient. For lymph node station labeling, 84.5% of lymph nodes are correctly assigned to their stations. Conclusions: Multiple-channel shape, intensity, and spatial prior features aggregated by a random forest classifier improve mediastinal lymph node detection on chest CT. Using the location information of segmented anatomic structures from the multiatlas formulation enables accurate identification of lymph node stations.« less

An assessment of the effectiveness of a random forest classifier for land-cover classification

NASA Astrophysics Data System (ADS)

Rodriguez-Galiano, V. F.; Ghimire, B.; Rogan, J.; Chica-Olmo, M.; Rigol-Sanchez, J. P.

2012-01-01

Land cover monitoring using remotely sensed data requires robust classification methods which allow for the accurate mapping of complex land cover and land use categories. Random forest (RF) is a powerful machine learning classifier that is relatively unknown in land remote sensing and has not been evaluated thoroughly by the remote sensing community compared to more conventional pattern recognition techniques. Key advantages of RF include: their non-parametric nature; high classification accuracy; and capability to determine variable importance. However, the split rules for classification are unknown, therefore RF can be considered to be black box type classifier. RF provides an algorithm for estimating missing values; and flexibility to perform several types of data analysis, including regression, classification, survival analysis, and unsupervised learning. In this paper, the performance of the RF classifier for land cover classification of a complex area is explored. Evaluation was based on several criteria: mapping accuracy, sensitivity to data set size and noise. Landsat-5 Thematic Mapper data captured in European spring and summer were used with auxiliary variables derived from a digital terrain model to classify 14 different land categories in the south of Spain. Results show that the RF algorithm yields accurate land cover classifications, with 92% overall accuracy and a Kappa index of 0.92. RF is robust to training data reduction and noise because significant differences in kappa values were only observed for data reduction and noise addition values greater than 50 and 20%, respectively. Additionally, variables that RF identified as most important for classifying land cover coincided with expectations. A McNemar test indicates an overall better performance of the random forest model over a single decision tree at the 0.00001 significance level.

Data-Driven Lead-Acid Battery Prognostics Using Random Survival Forests

DTIC Science & Technology

2014-10-02

Kogalur, Blackstone , & Lauer, 2008; Ishwaran & Kogalur, 2010). Random survival forest is a sur- vival analysis extension of Random Forests (Breiman, 2001...Statistics & probability letters, 80(13), 1056–1064. Ishwaran, H., Kogalur, U. B., Blackstone , E. H., & Lauer, M. S. (2008). Random survival forests. The

Estimation of Surface Seawater Fugacity of Carbon Dioxide Using Satellite Data and Machine Learning

NASA Astrophysics Data System (ADS)

Jang, E.; Im, J.; Park, G.; Park, Y.

2016-12-01

The ocean controls the climate of Earth by absorbing and releasing CO2 through the carbon cycle. The amount of CO2 in the ocean has increased since the industrial revolution. High CO2 concentration in the ocean has a negative influence to marine organisms and reduces the ability of absorbing CO2 in the ocean. This study estimated surface seawater fugacity of CO2 (fCO2) in the East Sea of Korea using Geostationary Ocean Color Imager (GOCI) and Moderate Resolution Imaging Spectroradiometer (MODIS) satellite data, and Hybrid Coordinate Ocean Model (HYCOM) reanalysis data. GOCI is the world first geostationary ocean color observation satellite sensor, and it provides 8 images with 8 bands hourly per day from 9 am to 4 pm at 500m resolution. Two machine learning approaches (i.e., random forest and support vector regression) were used to model fCO2 in this study. While most of the existing studies used multiple linear regression to estimate the pressure of CO2 in the ocean, machine learning may handle more complex relationship between surface seawater fCO2 and ocean parameters in a dynamic spatiotemporal environment. Five ocean related parameters, colored dissolved organic matter (CDOM), chlorophyll-a (chla), sea surface temperature (SST), sea surface salinity (SSS), and mixed layer depth (MLD), were used as input variables. This study examined two schemes, one with GOCI-derived products and the other with MODIS-derived ones. Results show that random forest performed better than support vector regression regardless of satellite data used. The accuracy of GOCI-based estimation was higher than MODIS-based one, possibly thanks to the better spatiotemporal resolution of GOCI data. MLD was identified the most contributing parameter in estimating surface seawater fCO2 among the five ocean related parameters, which might be related with an active deep convection in the East Sea. The surface seawater fCO2 in summer was higher in general with some spatial variation than the other seasons because of higher SST.

A machine learning approach for predicting methionine oxidation sites.

PubMed

Aledo, Juan C; Cantón, Francisco R; Veredas, Francisco J

2017-09-29

The oxidation of protein-bound methionine to form methionine sulfoxide, has traditionally been regarded as an oxidative damage. However, recent evidences support the view of this reversible reaction as a regulatory post-translational modification. The perception that methionine sulfoxidation may provide a mechanism to the redox regulation of a wide range of cellular processes, has stimulated some proteomic studies. However, these experimental approaches are expensive and time-consuming. Therefore, computational methods designed to predict methionine oxidation sites are an attractive alternative. As a first approach to this matter, we have developed models based on random forests, support vector machines and neural networks, aimed at accurate prediction of sites of methionine oxidation. Starting from published proteomic data regarding oxidized methionines, we created a hand-curated dataset formed by 113 unique polypeptides of known structure, containing 975 methionyl residues, 122 of which were oxidation-prone (positive dataset) and 853 were oxidation-resistant (negative dataset). We use a machine learning approach to generate predictive models from these datasets. Among the multiple features used in the classification task, some of them contributed substantially to the performance of the predictive models. Thus, (i) the solvent accessible area of the methionine residue, (ii) the number of residues between the analyzed methionine and the next methionine found towards the N-terminus and (iii) the spatial distance between the atom of sulfur from the analyzed methionine and the closest aromatic residue, were among the most relevant features. Compared to the other classifiers we also evaluated, random forests provided the best performance, with accuracy, sensitivity and specificity of 0.7468±0.0567, 0.6817±0.0982 and 0.7557±0.0721, respectively (mean ± standard deviation). We present the first predictive models aimed to computationally detect methionine sites that may become oxidized in vivo in response to oxidative signals. These models provide insights into the structural context in which a methionine residue become either oxidation-resistant or oxidation-prone. Furthermore, these models should be useful in prioritizing methinonyl residues for further studies to determine their potential as regulatory post-translational modification sites.

Simultaneous feature selection and parameter optimisation using an artificial ant colony: case study of melting point prediction.

PubMed

O'Boyle, Noel M; Palmer, David S; Nigsch, Florian; Mitchell, John Bo

2008-10-29

We present a novel feature selection algorithm, Winnowing Artificial Ant Colony (WAAC), that performs simultaneous feature selection and model parameter optimisation for the development of predictive quantitative structure-property relationship (QSPR) models. The WAAC algorithm is an extension of the modified ant colony algorithm of Shen et al. (J Chem Inf Model 2005, 45: 1024-1029). We test the ability of the algorithm to develop a predictive partial least squares model for the Karthikeyan dataset (J Chem Inf Model 2005, 45: 581-590) of melting point values. We also test its ability to perform feature selection on a support vector machine model for the same dataset. Starting from an initial set of 203 descriptors, the WAAC algorithm selected a PLS model with 68 descriptors which has an RMSE on an external test set of 46.6 degrees C and R2 of 0.51. The number of components chosen for the model was 49, which was close to optimal for this feature selection. The selected SVM model has 28 descriptors (cost of 5, epsilon of 0.21) and an RMSE of 45.1 degrees C and R2 of 0.54. This model outperforms a kNN model (RMSE of 48.3 degrees C, R2 of 0.47) for the same data and has similar performance to a Random Forest model (RMSE of 44.5 degrees C, R2 of 0.55). However it is much less prone to bias at the extremes of the range of melting points as shown by the slope of the line through the residuals: -0.43 for WAAC/SVM, -0.53 for Random Forest. With a careful choice of objective function, the WAAC algorithm can be used to optimise machine learning and regression models that suffer from overfitting. Where model parameters also need to be tuned, as is the case with support vector machine and partial least squares models, it can optimise these simultaneously. The moving probabilities used by the algorithm are easily interpreted in terms of the best and current models of the ants, and the winnowing procedure promotes the removal of irrelevant descriptors.

High-resolution digital mapping of soil organic carbon in permafrost terrain using machine learning: a case study in a sub-Arctic peatland environment

NASA Astrophysics Data System (ADS)

Siewert, Matthias B.

2018-03-01

Soil organic carbon (SOC) stored in northern peatlands and permafrost-affected soils are key components in the global carbon cycle. This article quantifies SOC stocks in a sub-Arctic mountainous peatland environment in the discontinuous permafrost zone in Abisko, northern Sweden. Four machine-learning techniques are evaluated for SOC quantification: multiple linear regression, artificial neural networks, support vector machine and random forest. The random forest model performed best and was used to predict SOC for several depth increments at a spatial resolution of 1 m (1×1 m). A high-resolution (1 m) land cover classification generated for this study is the most relevant predictive variable. The landscape mean SOC storage (0-150 cm) is estimated to be 8.3 ± 8.0 kg C m-2 and the SOC stored in the top meter (0-100 cm) to be 7.7 ± 6.2 kg C m-2. The predictive modeling highlights the relative importance of wetland areas and in particular peat plateaus for the landscape's SOC storage. The total SOC was also predicted at reduced spatial resolutions of 2, 10, 30, 100, 250 and 1000 m and shows a significant drop in land cover class detail and a tendency to underestimate the SOC at resolutions > 30 m. This is associated with the occurrence of many small-scale wetlands forming local hot-spots of SOC storage that are omitted at coarse resolutions. Sharp transitions in SOC storage associated with land cover and permafrost distribution are the most challenging methodological aspect. However, in this study, at local, regional and circum-Arctic scales, the main factor limiting robust SOC mapping efforts is the scarcity of soil pedon data from across the entire environmental space. For the Abisko region, past SOC and permafrost dynamics indicate that most of the SOC is barely 2000 years old and very dynamic. Future research needs to investigate the geomorphic response of permafrost degradation and the fate of SOC across all landscape compartments in post-permafrost landscapes.

Combining random forest and 2D correlation analysis to identify serum spectral signatures for neuro-oncology.

PubMed

Smith, Benjamin R; Ashton, Katherine M; Brodbelt, Andrew; Dawson, Timothy; Jenkinson, Michael D; Hunt, Neil T; Palmer, David S; Baker, Matthew J

2016-06-07

Fourier transform infrared (FTIR) spectroscopy has long been established as an analytical technique for the measurement of vibrational modes of molecular systems. More recently, FTIR has been used for the analysis of biofluids with the aim of becoming a tool to aid diagnosis. For the clinician, this represents a convenient, fast, non-subjective option for the study of biofluids and the diagnosis of disease states. The patient also benefits from this method, as the procedure for the collection of serum is much less invasive and stressful than traditional biopsy. This is especially true of patients in whom brain cancer is suspected. A brain biopsy is very unpleasant for the patient, potentially dangerous and can occasionally be inconclusive. We therefore present a method for the diagnosis of brain cancer from serum samples using FTIR and machine learning techniques. The scope of the study involved 433 patients from whom were collected 9 spectra each in the range 600-4000 cm(-1). To begin the development of the novel method, various pre-processing steps were investigated and ranked in terms of final accuracy of the diagnosis. Random forest machine learning was utilised as a classifier to separate patients into cancer or non-cancer categories based upon the intensities of wavenumbers present in their spectra. Generalised 2D correlational analysis was then employed to further augment the machine learning, and also to establish spectral features important for the distinction between cancer and non-cancer serum samples. Using these methods, sensitivities of up to 92.8% and specificities of up to 91.5% were possible. Furthermore, ratiometrics were also investigated in order to establish any correlations present in the dataset. We show a rapid, computationally light, accurate, statistically robust methodology for the identification of spectral features present in differing disease states. With current advances in IR technology, such as the development of rapid discrete frequency collection, this approach is of importance to enable future clinical translation and enables IR to achieve its potential.

VizieR Online Data Catalog: Gamma-ray AGN type determination (Hassan+, 2013)

NASA Astrophysics Data System (ADS)

Hassan, T.; Mirabal, N.; Contreras, J. L.; Oya, I.

2013-11-01

In this paper, we employ Support Vector Machines (SVMs) and Random Forest (RF) that embody two of the most robust supervised learning algorithms available today. We are interested in building classifiers that can distinguish between two AGN classes: BL Lacs and FSRQs. In the 2FGL, there is a total set of 1074 identified/associated AGN objects with the following labels: 'bzb' (BL Lacs), 'bzq' (FSRQs), 'agn' (other non-blazar AGN) and 'agu' (active galaxies of uncertain type). From this global set, we group the identified/associated blazars ('bzb' and 'bzq' labels) as the training/testing set of our algorithms. (2 data files).

Application of XGBoost algorithm in hourly PM2.5 concentration prediction

NASA Astrophysics Data System (ADS)

Pan, Bingyue

2018-02-01

In view of prediction techniques of hourly PM2.5 concentration in China, this paper applied the XGBoost(Extreme Gradient Boosting) algorithm to predict hourly PM2.5 concentration. The monitoring data of air quality in Tianjin city was analyzed by using XGBoost algorithm. The prediction performance of the XGBoost method is evaluated by comparing observed and predicted PM2.5 concentration using three measures of forecast accuracy. The XGBoost method is also compared with the random forest algorithm, multiple linear regression, decision tree regression and support vector machines for regression models using computational results. The results demonstrate that the XGBoost algorithm outperforms other data mining methods.

QUANTIFYING FOREST ABOVEGROUND CARBON POOLS AND FLUXES USING MULTI-TEMPORAL LIDAR A report on field monitoring, remote sensing MMV, GIS integration, and modeling results for forestry field validation test to quantify aboveground tree biomass and carbon

DOE Office of Scientific and Technical Information (OSTI.GOV)

Lee Spangler; Lee A. Vierling; Eva K. Stand

2012-04-01

Sound policy recommendations relating to the role of forest management in mitigating atmospheric carbon dioxide (CO{sub 2}) depend upon establishing accurate methodologies for quantifying forest carbon pools for large tracts of land that can be dynamically updated over time. Light Detection and Ranging (LiDAR) remote sensing is a promising technology for achieving accurate estimates of aboveground biomass and thereby carbon pools; however, not much is known about the accuracy of estimating biomass change and carbon flux from repeat LiDAR acquisitions containing different data sampling characteristics. In this study, discrete return airborne LiDAR data was collected in 2003 and 2009 acrossmore » {approx}20,000 hectares (ha) of an actively managed, mixed conifer forest landscape in northern Idaho, USA. Forest inventory plots, established via a random stratified sampling design, were established and sampled in 2003 and 2009. The Random Forest machine learning algorithm was used to establish statistical relationships between inventory data and forest structural metrics derived from the LiDAR acquisitions. Aboveground biomass maps were created for the study area based on statistical relationships developed at the plot level. Over this 6-year period, we found that the mean increase in biomass due to forest growth across the non-harvested portions of the study area was 4.8 metric ton/hectare (Mg/ha). In these non-harvested areas, we found a significant difference in biomass increase among forest successional stages, with a higher biomass increase in mature and old forest compared to stand initiation and young forest. Approximately 20% of the landscape had been disturbed by harvest activities during the six-year time period, representing a biomass loss of >70 Mg/ha in these areas. During the study period, these harvest activities outweighed growth at the landscape scale, resulting in an overall loss in aboveground carbon at this site. The 30-fold increase in sampling density between the 2003 and 2009 did not affect the biomass estimates. Overall, LiDAR data coupled with field reference data offer a powerful method for calculating pools and changes in aboveground carbon in forested systems. The results of our study suggest that multitemporal LiDAR-based approaches are likely to be useful for high quality estimates of aboveground carbon change in conifer forest systems.« less

Reverse engineering model structures for soil and ecosystem respiration: the potential of gene expression programming

NASA Astrophysics Data System (ADS)

Ilie, Iulia; Dittrich, Peter; Carvalhais, Nuno; Jung, Martin; Heinemeyer, Andreas; Migliavacca, Mirco; Morison, James I. L.; Sippel, Sebastian; Subke, Jens-Arne; Wilkinson, Matthew; Mahecha, Miguel D.

2017-09-01

Accurate model representation of land-atmosphere carbon fluxes is essential for climate projections. However, the exact responses of carbon cycle processes to climatic drivers often remain uncertain. Presently, knowledge derived from experiments, complemented by a steadily evolving body of mechanistic theory, provides the main basis for developing such models. The strongly increasing availability of measurements may facilitate new ways of identifying suitable model structures using machine learning. Here, we explore the potential of gene expression programming (GEP) to derive relevant model formulations based solely on the signals present in data by automatically applying various mathematical transformations to potential predictors and repeatedly evolving the resulting model structures. In contrast to most other machine learning regression techniques, the GEP approach generates readable models that allow for prediction and possibly for interpretation. Our study is based on two cases: artificially generated data and real observations. Simulations based on artificial data show that GEP is successful in identifying prescribed functions, with the prediction capacity of the models comparable to four state-of-the-art machine learning methods (random forests, support vector machines, artificial neural networks, and kernel ridge regressions). Based on real observations we explore the responses of the different components of terrestrial respiration at an oak forest in south-eastern England. We find that the GEP-retrieved models are often better in prediction than some established respiration models. Based on their structures, we find previously unconsidered exponential dependencies of respiration on seasonal ecosystem carbon assimilation and water dynamics. We noticed that the GEP models are only partly portable across respiration components, the identification of a general terrestrial respiration model possibly prevented by equifinality issues. Overall, GEP is a promising tool for uncovering new model structures for terrestrial ecology in the data-rich era, complementing more traditional modelling approaches.

Surface electromyography based muscle fatigue detection using high-resolution time-frequency methods and machine learning algorithms.

PubMed

Karthick, P A; Ghosh, Diptasree Maitra; Ramakrishnan, S

2018-02-01

Surface electromyography (sEMG) based muscle fatigue research is widely preferred in sports science and occupational/rehabilitation studies due to its noninvasiveness. However, these signals are complex, multicomponent and highly nonstationary with large inter-subject variations, particularly during dynamic contractions. Hence, time-frequency based machine learning methodologies can improve the design of automated system for these signals. In this work, the analysis based on high-resolution time-frequency methods, namely, Stockwell transform (S-transform), B-distribution (BD) and extended modified B-distribution (EMBD) are proposed to differentiate the dynamic muscle nonfatigue and fatigue conditions. The nonfatigue and fatigue segments of sEMG signals recorded from the biceps brachii of 52 healthy volunteers are preprocessed and subjected to S-transform, BD and EMBD. Twelve features are extracted from each method and prominent features are selected using genetic algorithm (GA) and binary particle swarm optimization (BPSO). Five machine learning algorithms, namely, naïve Bayes, support vector machine (SVM) of polynomial and radial basis kernel, random forest and rotation forests are used for the classification. The results show that all the proposed time-frequency distributions (TFDs) are able to show the nonstationary variations of sEMG signals. Most of the features exhibit statistically significant difference in the muscle fatigue and nonfatigue conditions. The maximum number of features (66%) is reduced by GA and BPSO for EMBD and BD-TFD respectively. The combination of EMBD- polynomial kernel based SVM is found to be most accurate (91% accuracy) in classifying the conditions with the features selected using GA. The proposed methods are found to be capable of handling the nonstationary and multicomponent variations of sEMG signals recorded in dynamic fatiguing contractions. Particularly, the combination of EMBD- polynomial kernel based SVM could be used to detect the dynamic muscle fatigue conditions. Copyright © 2017 Elsevier B.V. All rights reserved.

Agricultural land change in the Carpathian ecoregion after the breakdown of socialism and expansion of the European Union

NASA Astrophysics Data System (ADS)

Griffiths, Patrick; Müller, Daniel; Kuemmerle, Tobias; Hostert, Patrick

2013-12-01

Widespread changes of agricultural land use occurred in Eastern Europe since the collapse of socialism and the European Union’s eastward expansion, but the rates and patterns of recent land changes remain unclear. Here we assess agricultural land change for the entire Carpathian ecoregion in Eastern Europe at 30 m spatial resolution with Landsat data and for two change periods, between 1985-2000 and 2000-2010. The early period is characterized by post-socialist transition processes, the late period by an increasing influence of EU politics in the region. For mapping and change detection, we use a machine learning approach (random forests) on image composites and variance metrics which were derived from the full decadal archive of Landsat imagery. Our results suggest that cropland abandonment was the most prevalent change process, but we also detected considerable areas of grassland conversion and forest expansion on non-forest land. Cropland abandonment was most extensive during the transition period and predominantly occurred in marginal areas with low suitability for agriculture. Conversely, we observed substantial recultivation of formerly abandoned cropland in high-value agricultural areas since 2000. Hence, market forces increasingly adjust socialist legacies of land expansive production and agricultural land use clusters in favorable areas while marginal lands revert to forest.

Using machine learning to explore the long-term evolution of GRS 1915+105

NASA Astrophysics Data System (ADS)

Huppenkothen, Daniela; Heil, Lucy M.; Hogg, David W.; Mueller, Andreas

2017-04-01

Among the population of known Galactic black hole X-ray binaries, GRS 1915+105 stands out in multiple ways. It has been in continuous outburst since 1992, and has shown a wide range of different states that can be distinguished by their timing and spectral properties. These states, also observed in IGR J17091-3624, have in the past been linked to accretion dynamics. Here, we present the first comprehensive study into the long-term evolution of GRS 1915+105, using the entire data set observed with Rossi X-ray Timing Explorer over its 16-yr lifetime. We develop a set of descriptive features allowing for automatic separation of states, and show that supervised machine learning in the form of logistic regression and random forests can be used to efficiently classify the entire data set. For the first time, we explore the duty cycle and time evolution of states over the entire 16-yr time span, and find that the temporal distribution of states has likely changed over the span of the observations. We connect the machine classification with physical interpretations of the phenomenology in terms of chaotic and stochastic processes.

In silico prediction of Tetrahymena pyriformis toxicity for diverse industrial chemicals with substructure pattern recognition and machine learning methods.

PubMed

Cheng, Feixiong; Shen, Jie; Yu, Yue; Li, Weihua; Liu, Guixia; Lee, Philip W; Tang, Yun

2011-03-01

There is an increasing need for the rapid safety assessment of chemicals by both industries and regulatory agencies throughout the world. In silico techniques are practical alternatives in the environmental hazard assessment. It is especially true to address the persistence, bioaccumulative and toxicity potentials of organic chemicals. Tetrahymena pyriformis toxicity is often used as a toxic endpoint. In this study, 1571 diverse unique chemicals were collected from the literature and composed of the largest diverse data set for T. pyriformis toxicity. Classification predictive models of T. pyriformis toxicity were developed by substructure pattern recognition and different machine learning methods, including support vector machine (SVM), C4.5 decision tree, k-nearest neighbors and random forest. The results of a 5-fold cross-validation showed that the SVM method performed better than other algorithms. The overall predictive accuracies of the SVM classification model with radial basis functions kernel was 92.2% for the 5-fold cross-validation and 92.6% for the external validation set, respectively. Furthermore, several representative substructure patterns for characterizing T. pyriformis toxicity were also identified via the information gain analysis methods. Copyright © 2010 Elsevier Ltd. All rights reserved.

Assessing patient risk of central line-associated bacteremia via machine learning.

PubMed

Beeler, Cole; Dbeibo, Lana; Kelley, Kristen; Thatcher, Levi; Webb, Douglas; Bah, Amadou; Monahan, Patrick; Fowler, Nicole R; Nicol, Spencer; Judy-Malcolm, Alisa; Azar, Jose

2018-04-13

Central line-associated bloodstream infections (CLABSIs) contribute to increased morbidity, length of hospital stay, and cost. Despite progress in understanding the risk factors, there remains a need to accurately predict the risk of CLABSIs and, in real time, prevent them from occurring. A predictive model was developed using retrospective data from a large academic healthcare system. Models were developed with machine learning via construction of random forests using validated input variables. Fifteen variables accounted for the most significant effect on CLABSI prediction based on a retrospective study of 70,218 unique patient encounters between January 1, 2013, and May 31, 2016. The area under the receiver operating characteristic curve for the best-performing model was 0.82 in production. This model has multiple applications for resource allocation for CLABSI prevention, including serving as a tool to target patients at highest risk for potentially cost-effective but otherwise time-limited interventions. Machine learning can be used to develop accurate models to predict the risk of CLABSI in real time prior to the development of infection. Copyright © 2018 Association for Professionals in Infection Control and Epidemiology, Inc. Published by Elsevier Inc. All rights reserved.

Mental Health Risk Adjustment with Clinical Categories and Machine Learning.

PubMed

Shrestha, Akritee; Bergquist, Savannah; Montz, Ellen; Rose, Sherri

2017-12-15

To propose nonparametric ensemble machine learning for mental health and substance use disorders (MHSUD) spending risk adjustment formulas, including considering Clinical Classification Software (CCS) categories as diagnostic covariates over the commonly used Hierarchical Condition Category (HCC) system. 2012-2013 Truven MarketScan database. We implement 21 algorithms to predict MHSUD spending, as well as a weighted combination of these algorithms called super learning. The algorithm collection included seven unique algorithms that were supplied with three differing sets of MHSUD-related predictors alongside demographic covariates: HCC, CCS, and HCC + CCS diagnostic variables. Performance was evaluated based on cross-validated R 2 and predictive ratios. Results show that super learning had the best performance based on both metrics. The top single algorithm was random forests, which improved on ordinary least squares regression by 10 percent with respect to relative efficiency. CCS categories-based formulas were generally more predictive of MHSUD spending compared to HCC-based formulas. Literature supports the potential benefit of implementing a separate MHSUD spending risk adjustment formula. Our results suggest there is an incentive to explore machine learning for MHSUD-specific risk adjustment, as well as considering CCS categories over HCCs. © Health Research and Educational Trust.

Regulatory element-based prediction identifies new susceptibility regulatory variants for osteoporosis.

PubMed

Yao, Shi; Guo, Yan; Dong, Shan-Shan; Hao, Ruo-Han; Chen, Xiao-Feng; Chen, Yi-Xiao; Chen, Jia-Bin; Tian, Qing; Deng, Hong-Wen; Yang, Tie-Lin

2017-08-01

Despite genome-wide association studies (GWASs) have identified many susceptibility genes for osteoporosis, it still leaves a large part of missing heritability to be discovered. Integrating regulatory information and GWASs could offer new insights into the biological link between the susceptibility SNPs and osteoporosis. We generated five machine learning classifiers with osteoporosis-associated variants and regulatory features data. We gained the optimal classifier and predicted genome-wide SNPs to discover susceptibility regulatory variants. We further utilized Genetic Factors for Osteoporosis Consortium (GEFOS) and three in-house GWASs samples to validate the associations for predicted positive SNPs. The random forest classifier performed best among all machine learning methods with the F1 score of 0.8871. Using the optimized model, we predicted 37,584 candidate SNPs for osteoporosis. According to the meta-analysis results, a list of regulatory variants was significantly associated with osteoporosis after multiple testing corrections and contributed to the expression of known osteoporosis-associated protein-coding genes. In summary, combining GWASs and regulatory elements through machine learning could provide additional information for understanding the mechanism of osteoporosis. The regulatory variants we predicted will provide novel targets for etiology research and treatment of osteoporosis.

Deep Learning Accurately Predicts Estrogen Receptor Status in Breast Cancer Metabolomics Data.

PubMed

Alakwaa, Fadhl M; Chaudhary, Kumardeep; Garmire, Lana X

2018-01-05

Metabolomics holds the promise as a new technology to diagnose highly heterogeneous diseases. Conventionally, metabolomics data analysis for diagnosis is done using various statistical and machine learning based classification methods. However, it remains unknown if deep neural network, a class of increasingly popular machine learning methods, is suitable to classify metabolomics data. Here we use a cohort of 271 breast cancer tissues, 204 positive estrogen receptor (ER+), and 67 negative estrogen receptor (ER-) to test the accuracies of feed-forward networks, a deep learning (DL) framework, as well as six widely used machine learning models, namely random forest (RF), support vector machines (SVM), recursive partitioning and regression trees (RPART), linear discriminant analysis (LDA), prediction analysis for microarrays (PAM), and generalized boosted models (GBM). DL framework has the highest area under the curve (AUC) of 0.93 in classifying ER+/ER- patients, compared to the other six machine learning algorithms. Furthermore, the biological interpretation of the first hidden layer reveals eight commonly enriched significant metabolomics pathways (adjusted P-value <0.05) that cannot be discovered by other machine learning methods. Among them, protein digestion and absorption and ATP-binding cassette (ABC) transporters pathways are also confirmed in integrated analysis between metabolomics and gene expression data in these samples. In summary, deep learning method shows advantages for metabolomics based breast cancer ER status classification, with both the highest prediction accuracy (AUC = 0.93) and better revelation of disease biology. We encourage the adoption of feed-forward networks based deep learning method in the metabolomics research community for classification.

Seasonal streamflow forecast with machine learning and teleconnection indices in the context non-stationary climate

NASA Astrophysics Data System (ADS)

Haguma, D.; Leconte, R.

2017-12-01

Spatial and temporal water resources variability are associated with large-scale pressure and circulation anomalies known as teleconnections that influence the pattern of the atmospheric circulation. Teleconnection indices have been used successfully to forecast streamflow in short term. However, in some watersheds, classical methods cannot establish relationships between seasonal streamflow and teleconnection indices because of weak correlation. In this study, machine learning algorithms have been applied for seasonal streamflow forecast using teleconnection indices. Machine learning offers an alternative to classical methods to address the non-linear relationship between streamflow and teleconnection indices the context non-stationary climate. Two machine learning algorithms, random forest (RF) and support vector machine (SVM), with teleconnection indices associated with North American climatology, have been used to forecast inflows for one and two leading seasons for the Romaine River and Manicouagan River watersheds, located in Quebec, Canada. The indices are Pacific-North America (PNA), North Atlantic Oscillation (NAO), El Niño-Southern Oscillation (ENSO), Arctic Oscillation (AO) and Pacific Decadal Oscillation (PDO). The results showed that the machine learning algorithms have an important predictive power for seasonal streamflow for one and two leading seasons. The RF performed better for training and SVM generally have better results with high predictive capability for testing. The RF which is an ensemble method, allowed to assess the uncertainty of the forecast. The integration of teleconnection indices responds to the seasonal forecast of streamflow in the conditions of the non-stationarity the climate, although the teleconnection indices have a weak correlation with streamflow.

Osteoporosis risk prediction for bone mineral density assessment of postmenopausal women using machine learning.

PubMed

Yoo, Tae Keun; Kim, Sung Kean; Kim, Deok Won; Choi, Joon Yul; Lee, Wan Hyung; Oh, Ein; Park, Eun-Cheol

2013-11-01

A number of clinical decision tools for osteoporosis risk assessment have been developed to select postmenopausal women for the measurement of bone mineral density. We developed and validated machine learning models with the aim of more accurately identifying the risk of osteoporosis in postmenopausal women compared to the ability of conventional clinical decision tools. We collected medical records from Korean postmenopausal women based on the Korea National Health and Nutrition Examination Surveys. The training data set was used to construct models based on popular machine learning algorithms such as support vector machines (SVM), random forests, artificial neural networks (ANN), and logistic regression (LR) based on simple surveys. The machine learning models were compared to four conventional clinical decision tools: osteoporosis self-assessment tool (OST), osteoporosis risk assessment instrument (ORAI), simple calculated osteoporosis risk estimation (SCORE), and osteoporosis index of risk (OSIRIS). SVM had significantly better area under the curve (AUC) of the receiver operating characteristic than ANN, LR, OST, ORAI, SCORE, and OSIRIS for the training set. SVM predicted osteoporosis risk with an AUC of 0.827, accuracy of 76.7%, sensitivity of 77.8%, and specificity of 76.0% at total hip, femoral neck, or lumbar spine for the testing set. The significant factors selected by SVM were age, height, weight, body mass index, duration of menopause, duration of breast feeding, estrogen therapy, hyperlipidemia, hypertension, osteoarthritis, and diabetes mellitus. Considering various predictors associated with low bone density, the machine learning methods may be effective tools for identifying postmenopausal women at high risk for osteoporosis.

Machine-Learning Algorithms to Code Public Health Spending Accounts

PubMed Central

Leider, Jonathon P.; Resnick, Beth A.; Alfonso, Y. Natalia; Bishai, David

2017-01-01

Objectives: Government public health expenditure data sets require time- and labor-intensive manipulation to summarize results that public health policy makers can use. Our objective was to compare the performances of machine-learning algorithms with manual classification of public health expenditures to determine if machines could provide a faster, cheaper alternative to manual classification. Methods: We used machine-learning algorithms to replicate the process of manually classifying state public health expenditures, using the standardized public health spending categories from the Foundational Public Health Services model and a large data set from the US Census Bureau. We obtained a data set of 1.9 million individual expenditure items from 2000 to 2013. We collapsed these data into 147 280 summary expenditure records, and we followed a standardized method of manually classifying each expenditure record as public health, maybe public health, or not public health. We then trained 9 machine-learning algorithms to replicate the manual process. We calculated recall, precision, and coverage rates to measure the performance of individual and ensembled algorithms. Results: Compared with manual classification, the machine-learning random forests algorithm produced 84% recall and 91% precision. With algorithm ensembling, we achieved our target criterion of 90% recall by using a consensus ensemble of ≥6 algorithms while still retaining 93% coverage, leaving only 7% of the summary expenditure records unclassified. Conclusions: Machine learning can be a time- and cost-saving tool for estimating public health spending in the United States. It can be used with standardized public health spending categories based on the Foundational Public Health Services model to help parse public health expenditure information from other types of health-related spending, provide data that are more comparable across public health organizations, and evaluate the impact of evidence-based public health resource allocation. PMID:28363034

Machine learning approaches to the social determinants of health in the health and retirement study.

PubMed

Seligman, Benjamin; Tuljapurkar, Shripad; Rehkopf, David

2018-04-01

Social and economic factors are important predictors of health and of recognized importance for health systems. However, machine learning, used elsewhere in the biomedical literature, has not been extensively applied to study relationships between society and health. We investigate how machine learning may add to our understanding of social determinants of health using data from the Health and Retirement Study. A linear regression of age and gender, and a parsimonious theory-based regression additionally incorporating income, wealth, and education, were used to predict systolic blood pressure, body mass index, waist circumference, and telomere length. Prediction, fit, and interpretability were compared across four machine learning methods: linear regression, penalized regressions, random forests, and neural networks. All models had poor out-of-sample prediction. Most machine learning models performed similarly to the simpler models. However, neural networks greatly outperformed the three other methods. Neural networks also had good fit to the data ( R 2 between 0.4-0.6, versus <0.3 for all others). Across machine learning models, nine variables were frequently selected or highly weighted as predictors: dental visits, current smoking, self-rated health, serial-seven subtractions, probability of receiving an inheritance, probability of leaving an inheritance of at least $10,000, number of children ever born, African-American race, and gender. Some of the machine learning methods do not improve prediction or fit beyond simpler models, however, neural networks performed well. The predictors identified across models suggest underlying social factors that are important predictors of biological indicators of chronic disease, and that the non-linear and interactive relationships between variables fundamental to the neural network approach may be important to consider.

Machine-Learning Algorithms to Code Public Health Spending Accounts.

PubMed

Brady, Eoghan S; Leider, Jonathon P; Resnick, Beth A; Alfonso, Y Natalia; Bishai, David

Government public health expenditure data sets require time- and labor-intensive manipulation to summarize results that public health policy makers can use. Our objective was to compare the performances of machine-learning algorithms with manual classification of public health expenditures to determine if machines could provide a faster, cheaper alternative to manual classification. We used machine-learning algorithms to replicate the process of manually classifying state public health expenditures, using the standardized public health spending categories from the Foundational Public Health Services model and a large data set from the US Census Bureau. We obtained a data set of 1.9 million individual expenditure items from 2000 to 2013. We collapsed these data into 147 280 summary expenditure records, and we followed a standardized method of manually classifying each expenditure record as public health, maybe public health, or not public health. We then trained 9 machine-learning algorithms to replicate the manual process. We calculated recall, precision, and coverage rates to measure the performance of individual and ensembled algorithms. Compared with manual classification, the machine-learning random forests algorithm produced 84% recall and 91% precision. With algorithm ensembling, we achieved our target criterion of 90% recall by using a consensus ensemble of ≥6 algorithms while still retaining 93% coverage, leaving only 7% of the summary expenditure records unclassified. Machine learning can be a time- and cost-saving tool for estimating public health spending in the United States. It can be used with standardized public health spending categories based on the Foundational Public Health Services model to help parse public health expenditure information from other types of health-related spending, provide data that are more comparable across public health organizations, and evaluate the impact of evidence-based public health resource allocation.

Automatic classification of protein structures using physicochemical parameters.

PubMed

Mohan, Abhilash; Rao, M Divya; Sunderrajan, Shruthi; Pennathur, Gautam

2014-09-01

Protein classification is the first step to functional annotation; SCOP and Pfam databases are currently the most relevant protein classification schemes. However, the disproportion in the number of three dimensional (3D) protein structures generated versus their classification into relevant superfamilies/families emphasizes the need for automated classification schemes. Predicting function of novel proteins based on sequence information alone has proven to be a major challenge. The present study focuses on the use of physicochemical parameters in conjunction with machine learning algorithms (Naive Bayes, Decision Trees, Random Forest and Support Vector Machines) to classify proteins into their respective SCOP superfamily/Pfam family, using sequence derived information. Spectrophores™, a 1D descriptor of the 3D molecular field surrounding a structure was used as a benchmark to compare the performance of the physicochemical parameters. The machine learning algorithms were modified to select features based on information gain for each SCOP superfamily/Pfam family. The effect of combining physicochemical parameters and spectrophores on classification accuracy (CA) was studied. Machine learning algorithms trained with the physicochemical parameters consistently classified SCOP superfamilies and Pfam families with a classification accuracy above 90%, while spectrophores performed with a CA of around 85%. Feature selection improved classification accuracy for both physicochemical parameters and spectrophores based machine learning algorithms. Combining both attributes resulted in a marginal loss of performance. Physicochemical parameters were able to classify proteins from both schemes with classification accuracy ranging from 90-96%. These results suggest the usefulness of this method in classifying proteins from amino acid sequences.

Gradient boosting machine for modeling the energy consumption of commercial buildings

DOE PAGES

Touzani, Samir; Granderson, Jessica; Fernandes, Samuel

2017-11-26

Accurate savings estimations are important to promote energy efficiency projects and demonstrate their cost-effectiveness. The increasing presence of advanced metering infrastructure (AMI) in commercial buildings has resulted in a rising availability of high frequency interval data. These data can be used for a variety of energy efficiency applications such as demand response, fault detection and diagnosis, and heating, ventilation, and air conditioning (HVAC) optimization. This large amount of data has also opened the door to the use of advanced statistical learning models, which hold promise for providing accurate building baseline energy consumption predictions, and thus accurate saving estimations. The gradientmore » boosting machine is a powerful machine learning algorithm that is gaining considerable traction in a wide range of data driven applications, such as ecology, computer vision, and biology. In the present work an energy consumption baseline modeling method based on a gradient boosting machine was proposed. To assess the performance of this method, a recently published testing procedure was used on a large dataset of 410 commercial buildings. The model training periods were varied and several prediction accuracy metrics were used to evaluate the model's performance. The results show that using the gradient boosting machine model improved the R-squared prediction accuracy and the CV(RMSE) in more than 80 percent of the cases, when compared to an industry best practice model that is based on piecewise linear regression, and to a random forest algorithm.« less

Gradient boosting machine for modeling the energy consumption of commercial buildings

DOE Office of Scientific and Technical Information (OSTI.GOV)

Touzani, Samir; Granderson, Jessica; Fernandes, Samuel

Accurate savings estimations are important to promote energy efficiency projects and demonstrate their cost-effectiveness. The increasing presence of advanced metering infrastructure (AMI) in commercial buildings has resulted in a rising availability of high frequency interval data. These data can be used for a variety of energy efficiency applications such as demand response, fault detection and diagnosis, and heating, ventilation, and air conditioning (HVAC) optimization. This large amount of data has also opened the door to the use of advanced statistical learning models, which hold promise for providing accurate building baseline energy consumption predictions, and thus accurate saving estimations. The gradientmore » boosting machine is a powerful machine learning algorithm that is gaining considerable traction in a wide range of data driven applications, such as ecology, computer vision, and biology. In the present work an energy consumption baseline modeling method based on a gradient boosting machine was proposed. To assess the performance of this method, a recently published testing procedure was used on a large dataset of 410 commercial buildings. The model training periods were varied and several prediction accuracy metrics were used to evaluate the model's performance. The results show that using the gradient boosting machine model improved the R-squared prediction accuracy and the CV(RMSE) in more than 80 percent of the cases, when compared to an industry best practice model that is based on piecewise linear regression, and to a random forest algorithm.« less

A Machine Learning Approach to Automated Gait Analysis for the Noldus Catwalk System.

PubMed

Frohlich, Holger; Claes, Kasper; De Wolf, Catherine; Van Damme, Xavier; Michel, Anne

2018-05-01

Gait analysis of animal disease models can provide valuable insights into in vivo compound effects and thus help in preclinical drug development. The purpose of this paper is to establish a computational gait analysis approach for the Noldus Catwalk system, in which footprints are automatically captured and stored. We present a - to our knowledge - first machine learning based approach for the Catwalk system, which comprises a step decomposition, definition and extraction of meaningful features, multivariate step sequence alignment, feature selection, and training of different classifiers (gradient boosting machine, random forest, and elastic net). Using animal-wise leave-one-out cross validation we demonstrate that with our method we can reliable separate movement patterns of a putative Parkinson's disease animal model and several control groups. Furthermore, we show that we can predict the time point after and the type of different brain lesions and can even forecast the brain region, where the intervention was applied. We provide an in-depth analysis of the features involved into our classifiers via statistical techniques for model interpretation. A machine learning method for automated analysis of data from the Noldus Catwalk system was established. Our works shows the ability of machine learning to discriminate pharmacologically relevant animal groups based on their walking behavior in a multivariate manner. Further interesting aspects of the approach include the ability to learn from past experiments, improve with more data arriving and to make predictions for single animals in future studies.

Machine Learning Approach to Optimizing Combined Stimulation and Medication Therapies for Parkinson's Disease.

PubMed

Shamir, Reuben R; Dolber, Trygve; Noecker, Angela M; Walter, Benjamin L; McIntyre, Cameron C

2015-01-01

Deep brain stimulation (DBS) of the subthalamic region is an established therapy for advanced Parkinson's disease (PD). However, patients often require time-intensive post-operative management to balance their coupled stimulation and medication treatments. Given the large and complex parameter space associated with this task, we propose that clinical decision support systems (CDSS) based on machine learning algorithms could assist in treatment optimization. Develop a proof-of-concept implementation of a CDSS that incorporates patient-specific details on both stimulation and medication. Clinical data from 10 patients, and 89 post-DBS surgery visits, were used to create a prototype CDSS. The system was designed to provide three key functions: (1) information retrieval; (2) visualization of treatment, and; (3) recommendation on expected effective stimulation and drug dosages, based on three machine learning methods that included support vector machines, Naïve Bayes, and random forest. Measures of medication dosages, time factors, and symptom-specific pre-operative response to levodopa were significantly correlated with post-operative outcomes (P < 0.05) and their effect on outcomes was of similar magnitude to that of DBS. Using those results, the combined machine learning algorithms were able to accurately predict 86% (12/14) of the motor improvement scores at one year after surgery. Using patient-specific details, an appropriately parameterized CDSS could help select theoretically optimal DBS parameter settings and medication dosages that have potential to improve the clinical management of PD patients. Copyright © 2015 Elsevier Inc. All rights reserved.

THE MILKY WAY PROJECT: LEVERAGING CITIZEN SCIENCE AND MACHINE LEARNING TO DETECT INTERSTELLAR BUBBLES

DOE Office of Scientific and Technical Information (OSTI.GOV)

Beaumont, Christopher N.; Williams, Jonathan P.; Goodman, Alyssa A.

We present Brut, an algorithm to identify bubbles in infrared images of the Galactic midplane. Brut is based on the Random Forest algorithm, and uses bubbles identified by >35,000 citizen scientists from the Milky Way Project to discover the identifying characteristics of bubbles in images from the Spitzer Space Telescope. We demonstrate that Brut's ability to identify bubbles is comparable to expert astronomers. We use Brut to re-assess the bubbles in the Milky Way Project catalog, and find that 10%-30% of the objects in this catalog are non-bubble interlopers. Relative to these interlopers, high-reliability bubbles are more confined to themore » mid-plane, and display a stronger excess of young stellar objects along and within bubble rims. Furthermore, Brut is able to discover bubbles missed by previous searches—particularly bubbles near bright sources which have low contrast relative to their surroundings. Brut demonstrates the synergies that exist between citizen scientists, professional scientists, and machine learning techniques. In cases where ''untrained' citizens can identify patterns that machines cannot detect without training, machine learning algorithms like Brut can use the output of citizen science projects as input training sets, offering tremendous opportunities to speed the pace of scientific discovery. A hybrid model of machine learning combined with crowdsourced training data from citizen scientists can not only classify large quantities of data, but also address the weakness of each approach if deployed alone.« less

A finite element-based machine learning approach for modeling the mechanical behavior of the breast tissues under compression in real-time.

PubMed

Martínez-Martínez, F; Rupérez-Moreno, M J; Martínez-Sober, M; Solves-Llorens, J A; Lorente, D; Serrano-López, A J; Martínez-Sanchis, S; Monserrat, C; Martín-Guerrero, J D

2017-11-01

This work presents a data-driven method to simulate, in real-time, the biomechanical behavior of the breast tissues in some image-guided interventions such as biopsies or radiotherapy dose delivery as well as to speed up multimodal registration algorithms. Ten real breasts were used for this work. Their deformation due to the displacement of two compression plates was simulated off-line using the finite element (FE) method. Three machine learning models were trained with the data from those simulations. Then, they were used to predict in real-time the deformation of the breast tissues during the compression. The models were a decision tree and two tree-based ensemble methods (extremely randomized trees and random forest). Two different experimental setups were designed to validate and study the performance of these models under different conditions. The mean 3D Euclidean distance between nodes predicted by the models and those extracted from the FE simulations was calculated to assess the performance of the models in the validation set. The experiments proved that extremely randomized trees performed better than the other two models. The mean error committed by the three models in the prediction of the nodal displacements was under 2 mm, a threshold usually set for clinical applications. The time needed for breast compression prediction is sufficiently short to allow its use in real-time (<0.2 s). Copyright © 2017 Elsevier Ltd. All rights reserved.

Low-Quality Structural and Interaction Data Improves Binding Affinity Prediction via Random Forest.

PubMed

Li, Hongjian; Leung, Kwong-Sak; Wong, Man-Hon; Ballester, Pedro J

2015-06-12

Docking scoring functions can be used to predict the strength of protein-ligand binding. It is widely believed that training a scoring function with low-quality data is detrimental for its predictive performance. Nevertheless, there is a surprising lack of systematic validation experiments in support of this hypothesis. In this study, we investigated to which extent training a scoring function with data containing low-quality structural and binding data is detrimental for predictive performance. We actually found that low-quality data is not only non-detrimental, but beneficial for the predictive performance of machine-learning scoring functions, though the improvement is less important than that coming from high-quality data. Furthermore, we observed that classical scoring functions are not able to effectively exploit data beyond an early threshold, regardless of its quality. This demonstrates that exploiting a larger data volume is more important for the performance of machine-learning scoring functions than restricting to a smaller set of higher data quality.

Materials Screening for the Discovery of New Half-Heuslers: Machine Learning versus ab Initio Methods.

PubMed

Legrain, Fleur; Carrete, Jesús; van Roekeghem, Ambroise; Madsen, Georg K H; Mingo, Natalio

2018-01-18

Machine learning (ML) is increasingly becoming a helpful tool in the search for novel functional compounds. Here we use classification via random forests to predict the stability of half-Heusler (HH) compounds, using only experimentally reported compounds as a training set. Cross-validation yields an excellent agreement between the fraction of compounds classified as stable and the actual fraction of truly stable compounds in the ICSD. The ML model is then employed to screen 71 178 different 1:1:1 compositions, yielding 481 likely stable candidates. The predicted stability of HH compounds from three previous high-throughput ab initio studies is critically analyzed from the perspective of the alternative ML approach. The incomplete consistency among the three separate ab initio studies and between them and the ML predictions suggests that additional factors beyond those considered by ab initio phase stability calculations might be determinant to the stability of the compounds. Such factors can include configurational entropies and quasiharmonic contributions.

Cooperative photometric redshift estimation

NASA Astrophysics Data System (ADS)

Cavuoti, S.; Tortora, C.; Brescia, M.; Longo, G.; Radovich, M.; Napolitano, N. R.; Amaro, V.; Vellucci, C.

2017-06-01

In the modern galaxy surveys photometric redshifts play a central role in a broad range of studies, from gravitational lensing and dark matter distribution to galaxy evolution. Using a dataset of ~ 25,000 galaxies from the second data release of the Kilo Degree Survey (KiDS) we obtain photometric redshifts with five different methods: (i) Random forest, (ii) Multi Layer Perceptron with Quasi Newton Algorithm, (iii) Multi Layer Perceptron with an optimization network based on the Levenberg-Marquardt learning rule, (iv) the Bayesian Photometric Redshift model (or BPZ) and (v) a classical SED template fitting procedure (Le Phare). We show how SED fitting techniques could provide useful information on the galaxy spectral type which can be used to improve the capability of machine learning methods constraining systematic errors and reduce the occurrence of catastrophic outliers. We use such classification to train specialized regression estimators, by demonstrating that such hybrid approach, involving SED fitting and machine learning in a single collaborative framework, is capable to improve the overall prediction accuracy of photometric redshifts.

Classification Algorithms for Big Data Analysis, a Map Reduce Approach

NASA Astrophysics Data System (ADS)

Ayma, V. A.; Ferreira, R. S.; Happ, P.; Oliveira, D.; Feitosa, R.; Costa, G.; Plaza, A.; Gamba, P.

2015-03-01

Since many years ago, the scientific community is concerned about how to increase the accuracy of different classification methods, and major achievements have been made so far. Besides this issue, the increasing amount of data that is being generated every day by remote sensors raises more challenges to be overcome. In this work, a tool within the scope of InterIMAGE Cloud Platform (ICP), which is an open-source, distributed framework for automatic image interpretation, is presented. The tool, named ICP: Data Mining Package, is able to perform supervised classification procedures on huge amounts of data, usually referred as big data, on a distributed infrastructure using Hadoop MapReduce. The tool has four classification algorithms implemented, taken from WEKA's machine learning library, namely: Decision Trees, Naïve Bayes, Random Forest and Support Vector Machines (SVM). The results of an experimental analysis using a SVM classifier on data sets of different sizes for different cluster configurations demonstrates the potential of the tool, as well as aspects that affect its performance.

The prediction of food additives in the fruit juice based on electronic nose with chemometrics.

PubMed

Qiu, Shanshan; Wang, Jun

2017-09-01

Food additives are added to products to enhance their taste, and preserve flavor or appearance. While their use should be restricted to achieve a technological benefit, the contents of food additives should be also strictly controlled. In this study, E-nose was applied as an alternative to traditional monitoring technologies for determining two food additives, namely benzoic acid and chitosan. For quantitative monitoring, support vector machine (SVM), random forest (RF), extreme learning machine (ELM) and partial least squares regression (PLSR) were applied to establish regression models between E-nose signals and the amount of food additives in fruit juices. The monitoring models based on ELM and RF reached higher correlation coefficients (R 2 s) and lower root mean square errors (RMSEs) than models based on PLSR and SVM. This work indicates that E-nose combined with RF or ELM can be a cost-effective, easy-to-build and rapid detection system for food additive monitoring. Copyright © 2017 Elsevier Ltd. All rights reserved.

Mapping forested wetlands in the Great Zhan River Basin through integrating optical, radar, and topographical data classification techniques.

PubMed

Na, X D; Zang, S Y; Wu, C S; Li, W L

2015-11-01

Knowledge of the spatial extent of forested wetlands is essential to many studies including wetland functioning assessment, greenhouse gas flux estimation, and wildlife suitable habitat identification. For discriminating forested wetlands from their adjacent land cover types, researchers have resorted to image analysis techniques applied to numerous remotely sensed data. While with some success, there is still no consensus on the optimal approaches for mapping forested wetlands. To address this problem, we examined two machine learning approaches, random forest (RF) and K-nearest neighbor (KNN) algorithms, and applied these two approaches to the framework of pixel-based and object-based classifications. The RF and KNN algorithms were constructed using predictors derived from Landsat 8 imagery, Radarsat-2 advanced synthetic aperture radar (SAR), and topographical indices. The results show that the objected-based classifications performed better than per-pixel classifications using the same algorithm (RF) in terms of overall accuracy and the difference of their kappa coefficients are statistically significant (p<0.01). There were noticeably omissions for forested and herbaceous wetlands based on the per-pixel classifications using the RF algorithm. As for the object-based image analysis, there were also statistically significant differences (p<0.01) of Kappa coefficient between results performed based on RF and KNN algorithms. The object-based classification using RF provided a more visually adequate distribution of interested land cover types, while the object classifications based on the KNN algorithm showed noticeably commissions for forested wetlands and omissions for agriculture land. This research proves that the object-based classification with RF using optical, radar, and topographical data improved the mapping accuracy of land covers and provided a feasible approach to discriminate the forested wetlands from the other land cover types in forestry area.

Advantages of Synthetic Noise and Machine Learning for Analyzing Radioecological Data Sets.

PubMed

Shuryak, Igor

2017-01-01

The ecological effects of accidental or malicious radioactive contamination are insufficiently understood because of the hazards and difficulties associated with conducting studies in radioactively-polluted areas. Data sets from severely contaminated locations can therefore be small. Moreover, many potentially important factors, such as soil concentrations of toxic chemicals, pH, and temperature, can be correlated with radiation levels and with each other. In such situations, commonly-used statistical techniques like generalized linear models (GLMs) may not be able to provide useful information about how radiation and/or these other variables affect the outcome (e.g. abundance of the studied organisms). Ensemble machine learning methods such as random forests offer powerful alternatives. We propose that analysis of small radioecological data sets by GLMs and/or machine learning can be made more informative by using the following techniques: (1) adding synthetic noise variables to provide benchmarks for distinguishing the performances of valuable predictors from irrelevant ones; (2) adding noise directly to the predictors and/or to the outcome to test the robustness of analysis results against random data fluctuations; (3) adding artificial effects to selected predictors to test the sensitivity of the analysis methods in detecting predictor effects; (4) running a selected machine learning method multiple times (with different random-number seeds) to test the robustness of the detected "signal"; (5) using several machine learning methods to test the "signal's" sensitivity to differences in analysis techniques. Here, we applied these approaches to simulated data, and to two published examples of small radioecological data sets: (I) counts of fungal taxa in samples of soil contaminated by the Chernobyl nuclear power plan accident (Ukraine), and (II) bacterial abundance in soil samples under a ruptured nuclear waste storage tank (USA). We show that the proposed techniques were advantageous compared with the methodology used in the original publications where the data sets were presented. Specifically, our approach identified a negative effect of radioactive contamination in data set I, and suggested that in data set II stable chromium could have been a stronger limiting factor for bacterial abundance than the radionuclides 137Cs and 99Tc. This new information, which was extracted from these data sets using the proposed techniques, can potentially enhance the design of radioactive waste bioremediation.

Advantages of Synthetic Noise and Machine Learning for Analyzing Radioecological Data Sets

PubMed Central

Shuryak, Igor

2017-01-01

The ecological effects of accidental or malicious radioactive contamination are insufficiently understood because of the hazards and difficulties associated with conducting studies in radioactively-polluted areas. Data sets from severely contaminated locations can therefore be small. Moreover, many potentially important factors, such as soil concentrations of toxic chemicals, pH, and temperature, can be correlated with radiation levels and with each other. In such situations, commonly-used statistical techniques like generalized linear models (GLMs) may not be able to provide useful information about how radiation and/or these other variables affect the outcome (e.g. abundance of the studied organisms). Ensemble machine learning methods such as random forests offer powerful alternatives. We propose that analysis of small radioecological data sets by GLMs and/or machine learning can be made more informative by using the following techniques: (1) adding synthetic noise variables to provide benchmarks for distinguishing the performances of valuable predictors from irrelevant ones; (2) adding noise directly to the predictors and/or to the outcome to test the robustness of analysis results against random data fluctuations; (3) adding artificial effects to selected predictors to test the sensitivity of the analysis methods in detecting predictor effects; (4) running a selected machine learning method multiple times (with different random-number seeds) to test the robustness of the detected “signal”; (5) using several machine learning methods to test the “signal’s” sensitivity to differences in analysis techniques. Here, we applied these approaches to simulated data, and to two published examples of small radioecological data sets: (I) counts of fungal taxa in samples of soil contaminated by the Chernobyl nuclear power plan accident (Ukraine), and (II) bacterial abundance in soil samples under a ruptured nuclear waste storage tank (USA). We show that the proposed techniques were advantageous compared with the methodology used in the original publications where the data sets were presented. Specifically, our approach identified a negative effect of radioactive contamination in data set I, and suggested that in data set II stable chromium could have been a stronger limiting factor for bacterial abundance than the radionuclides 137Cs and 99Tc. This new information, which was extracted from these data sets using the proposed techniques, can potentially enhance the design of radioactive waste bioremediation. PMID:28068401

A hybrid training approach for leaf area index estimation via Cubist and random forests machine-learning

NASA Astrophysics Data System (ADS)

Houborg, Rasmus; McCabe, Matthew F.

2018-01-01

With an increasing volume and dimensionality of Earth observation data, enhanced integration of machine-learning methodologies is needed to effectively analyze and utilize these information rich datasets. In machine-learning, a training dataset is required to establish explicit associations between a suite of explanatory 'predictor' variables and the target property. The specifics of this learning process can significantly influence model validity and portability, with a higher generalization level expected with an increasing number of observable conditions being reflected in the training dataset. Here we propose a hybrid training approach for leaf area index (LAI) estimation, which harnesses synergistic attributes of scattered in-situ measurements and systematically distributed physically based model inversion results to enhance the information content and spatial representativeness of the training data. To do this, a complimentary training dataset of independent LAI was derived from a regularized model inversion of RapidEye surface reflectances and subsequently used to guide the development of LAI regression models via Cubist and random forests (RF) decision tree methods. The application of the hybrid training approach to a broad set of Landsat 8 vegetation index (VI) predictor variables resulted in significantly improved LAI prediction accuracies and spatial consistencies, relative to results relying on in-situ measurements alone for model training. In comparing the prediction capacity and portability of the two machine-learning algorithms, a pair of relatively simple multi-variate regression models established by Cubist performed best, with an overall relative mean absolute deviation (rMAD) of ∼11%, determined based on a stringent scene-specific cross-validation approach. In comparison, the portability of RF regression models was less effective (i.e., an overall rMAD of ∼15%), which was attributed partly to model saturation at high LAI in association with inherent extrapolation and transferability limitations. Explanatory VIs formed from bands in the near-infrared (NIR) and shortwave infrared domains (e.g., NDWI) were associated with the highest predictive ability, whereas Cubist models relying entirely on VIs based on NIR and red band combinations (e.g., NDVI) were associated with comparatively high uncertainties (i.e., rMAD ∼ 21%). The most transferable and best performing models were based on combinations of several predictor variables, which included both NDWI- and NDVI-like variables. In this process, prior screening of input VIs based on an assessment of variable relevance served as an effective mechanism for optimizing prediction accuracies from both Cubist and RF. While this study demonstrated benefit in combining data mining operations with physically based constraints via a hybrid training approach, the concept of transferability and portability warrants further investigations in order to realize the full potential of emerging machine-learning techniques for regression purposes.

Using machine learning to examine medication adherence thresholds and risk of hospitalization.

PubMed

Lo-Ciganic, Wei-Hsuan; Donohue, Julie M; Thorpe, Joshua M; Perera, Subashan; Thorpe, Carolyn T; Marcum, Zachary A; Gellad, Walid F

2015-08-01

Quality improvement efforts are frequently tied to patients achieving ≥80% medication adherence. However, there is little empirical evidence that this threshold optimally predicts important health outcomes. To apply machine learning to examine how adherence to oral hypoglycemic medications is associated with avoidance of hospitalizations, and to identify adherence thresholds for optimal discrimination of hospitalization risk. A retrospective cohort study of 33,130 non-dual-eligible Medicaid enrollees with type 2 diabetes. We randomly selected 90% of the cohort (training sample) to develop the prediction algorithm and used the remaining (testing sample) for validation. We applied random survival forests to identify predictors for hospitalization and fit survival trees to empirically derive adherence thresholds that best discriminate hospitalization risk, using the proportion of days covered (PDC). Time to first all-cause and diabetes-related hospitalization. The training and testing samples had similar characteristics (mean age, 48 y; 67% female; mean PDC=0.65). We identified 8 important predictors of all-cause hospitalizations (rank in order): prior hospitalizations/emergency department visit, number of prescriptions, diabetes complications, insulin use, PDC, number of prescribers, Elixhauser index, and eligibility category. The adherence thresholds most discriminating for risk of all-cause hospitalization varied from 46% to 94% according to patient health and medication complexity. PDC was not predictive of hospitalizations in the healthiest or most complex patient subgroups. Adherence thresholds most discriminating of hospitalization risk were not uniformly 80%. Machine-learning approaches may be valuable to identify appropriate patient-specific adherence thresholds for measuring quality of care and targeting nonadherent patients for intervention.

Brainstorming: weighted voting prediction of inhibitors for protein targets.

PubMed

Plewczynski, Dariusz

2011-09-01

The "Brainstorming" approach presented in this paper is a weighted voting method that can improve the quality of predictions generated by several machine learning (ML) methods. First, an ensemble of heterogeneous ML algorithms is trained on available experimental data, then all solutions are gathered and a consensus is built between them. The final prediction is performed using a voting procedure, whereby the vote of each method is weighted according to a quality coefficient calculated using multivariable linear regression (MLR). The MLR optimization procedure is very fast, therefore no additional computational cost is introduced by using this jury approach. Here, brainstorming is applied to selecting actives from large collections of compounds relating to five diverse biological targets of medicinal interest, namely HIV-reverse transcriptase, cyclooxygenase-2, dihydrofolate reductase, estrogen receptor, and thrombin. The MDL Drug Data Report (MDDR) database was used for selecting known inhibitors for these protein targets, and experimental data was then used to train a set of machine learning methods. The benchmark dataset (available at http://bio.icm.edu.pl/∼darman/chemoinfo/benchmark.tar.gz ) can be used for further testing of various clustering and machine learning methods when predicting the biological activity of compounds. Depending on the protein target, the overall recall value is raised by at least 20% in comparison to any single machine learning method (including ensemble methods like random forest) and unweighted simple majority voting procedures.

Engine classification using vibrations measured by Laser Doppler Vibrometer on different surfaces

NASA Astrophysics Data System (ADS)

Wei, J.; Liu, Chi-Him; Zhu, Zhigang; Vongsy, Karmon; Mendoza-Schrock, Olga

2015-05-01

In our previous studies, vehicle surfaces' vibrations caused by operating engines measured by Laser Doppler Vibrometer (LDV) have been effectively exploited in order to classify vehicles of different types, e.g., vans, 2-door sedans, 4-door sedans, trucks, and buses, as well as different types of engines, such as Inline-four engines, V-6 engines, 1-axle diesel engines, and 2-axle diesel engines. The results are achieved by employing methods based on an array of machine learning classifiers such as AdaBoost, random forests, neural network, and support vector machines. To achieve effective classification performance, we seek to find a more reliable approach to pick authentic vibrations of vehicle engines from a trustworthy surface. Compared with vibrations directly taken from the uncooperative vehicle surfaces that are rigidly connected to the engines, these vibrations are much weaker in magnitudes. In this work we conducted a systematic study on different types of objects. We tested different types of engines ranging from electric shavers, electric fans, and coffee machines among different surfaces such as a white board, cement wall, and steel case to investigate the characteristics of the LDV signals of these surfaces, in both the time and spectral domains. Preliminary results in engine classification using several machine learning algorithms point to the right direction on the choice of type of object surfaces to be planted for LDV measurements.

Radiomic machine-learning classifiers for prognostic biomarkers of advanced nasopharyngeal carcinoma.

PubMed

Zhang, Bin; He, Xin; Ouyang, Fusheng; Gu, Dongsheng; Dong, Yuhao; Zhang, Lu; Mo, Xiaokai; Huang, Wenhui; Tian, Jie; Zhang, Shuixing

2017-09-10

We aimed to identify optimal machine-learning methods for radiomics-based prediction of local failure and distant failure in advanced nasopharyngeal carcinoma (NPC). We enrolled 110 patients with advanced NPC. A total of 970 radiomic features were extracted from MRI images for each patient. Six feature selection methods and nine classification methods were evaluated in terms of their performance. We applied the 10-fold cross-validation as the criterion for feature selection and classification. We repeated each combination for 50 times to obtain the mean area under the curve (AUC) and test error. We observed that the combination methods Random Forest (RF) + RF (AUC, 0.8464 ± 0.0069; test error, 0.3135 ± 0.0088) had the highest prognostic performance, followed by RF + Adaptive Boosting (AdaBoost) (AUC, 0.8204 ± 0.0095; test error, 0.3384 ± 0.0097), and Sure Independence Screening (SIS) + Linear Support Vector Machines (LSVM) (AUC, 0.7883 ± 0.0096; test error, 0.3985 ± 0.0100). Our radiomics study identified optimal machine-learning methods for the radiomics-based prediction of local failure and distant failure in advanced NPC, which could enhance the applications of radiomics in precision oncology and clinical practice. Copyright © 2017 Elsevier B.V. All rights reserved.

Model-based and Model-free Machine Learning Techniques for Diagnostic Prediction and Classification of Clinical Outcomes in Parkinson's Disease.

PubMed

Gao, Chao; Sun, Hanbo; Wang, Tuo; Tang, Ming; Bohnen, Nicolaas I; Müller, Martijn L T M; Herman, Talia; Giladi, Nir; Kalinin, Alexandr; Spino, Cathie; Dauer, William; Hausdorff, Jeffrey M; Dinov, Ivo D

2018-05-08

In this study, we apply a multidisciplinary approach to investigate falls in PD patients using clinical, demographic and neuroimaging data from two independent initiatives (University of Michigan and Tel Aviv Sourasky Medical Center). Using machine learning techniques, we construct predictive models to discriminate fallers and non-fallers. Through controlled feature selection, we identified the most salient predictors of patient falls including gait speed, Hoehn and Yahr stage, postural instability and gait difficulty-related measurements. The model-based and model-free analytical methods we employed included logistic regression, random forests, support vector machines, and XGboost. The reliability of the forecasts was assessed by internal statistical (5-fold) cross validation as well as by external out-of-bag validation. Four specific challenges were addressed in the study: Challenge 1, develop a protocol for harmonizing and aggregating complex, multisource, and multi-site Parkinson's disease data; Challenge 2, identify salient predictive features associated with specific clinical traits, e.g., patient falls; Challenge 3, forecast patient falls and evaluate the classification performance; and Challenge 4, predict tremor dominance (TD) vs. posture instability and gait difficulty (PIGD). Our findings suggest that, compared to other approaches, model-free machine learning based techniques provide a more reliable clinical outcome forecasting of falls in Parkinson's patients, for example, with a classification accuracy of about 70-80%.

Estimating Ground-Level Particulate Matter (PM) Concentration using Satellite-derived Aerosol Optical Depth (AOD)

NASA Astrophysics Data System (ADS)

Park, Seohui; Im, Jungho

2017-04-01

Atmospheric aerosols are strongly associated with adverse human health effects. In particular, particulate matter less than 10 micrometers and 2.5 micrometers (i.e., PM10 and PM2.5, respectively) can cause cardiovascular and lung diseases such as asthma and chronic obstructive pulmonary disease (COPD). Air quality including PM has typically been monitored using station-based in-situ measurements over the world. However, in situ measurements do not provide spatial continuity over large areas. An alternative approach is to use satellite remote sensing as it provides data over vast areas at high temporal resolution. The literature shows that PM concentrations are related with Aerosol Optical Depth (AOD) that is derived from satellite observations, but it is still difficult to identify PM concentrations directly from AOD. Some studies used statistical approaches for estimating PM concentrations from AOD while some others combined numerical models and satellite-derived AOD. In this study, satellite-derived products were used to estimate ground PM concentrations based on machine learning over South Korea. Satellite-derived products include AOD from Geostationary Ocean Color Imager (GOCI), precipitation from Tropical Rainfall Measuring Mission (TRMM), soil moisture from AMSR-2, elevation from Shuttle Radar Topography Mission (SRTM), and land cover, land surface temperature and normalized difference vegetation index (NDVI) from Moderate Resolution Imaging Spectroradiometer (MODIS). PM concentrations data were collected from 318 stations. A statistical ordinary least squares (OLS) approach was also tested and compared with the machine learning approach (i.e., random forest). PM concentration was estimated during spring season (from March to May) in 2015 that typically shows high concentration of PM. The randomly selected 80% of data were used for model calibration and the remaining 20% were used for validation. The developed models were further tested for prediction of PM concentration. Results show that the estimation of PM10 was better than that of PM2.5 for both approaches. The performance of machine learning random forest was better (R2=0.53 and RMSE=17.74µm/m3 for PM10; R2=0.36 and RMSE=26.17 µm/m3 for PM2.5) than the statistical OLS approach (R2=0.13 and RMSE=23.66µm/m3 for PM10; R2=0.09 and RMSE=27.74 µm/m3 for PM2.5). However, both approaches did not fully model the entire dynamic range of PM concentrations, especially for very high concentrations, resulting in moderate underestimation.

Building rooftop classification using random forests for large-scale PV deployment

NASA Astrophysics Data System (ADS)

Assouline, Dan; Mohajeri, Nahid; Scartezzini, Jean-Louis

2017-10-01

Large scale solar Photovoltaic (PV) deployment on existing building rooftops has proven to be one of the most efficient and viable sources of renewable energy in urban areas. As it usually requires a potential analysis over the area of interest, a crucial step is to estimate the geometric characteristics of the building rooftops. In this paper, we introduce a multi-layer machine learning methodology to classify 6 roof types, 9 aspect (azimuth) classes and 5 slope (tilt) classes for all building rooftops in Switzerland, using GIS processing. We train Random Forests (RF), an ensemble learning algorithm, to build the classifiers. We use (2 × 2) [m2 ] LiDAR data (considering buildings and vegetation) to extract several rooftop features, and a generalised footprint polygon data to localize buildings. The roof classifier is trained and tested with 1252 labeled roofs from three different urban areas, namely Baden, Luzern, and Winterthur. The results for roof type classification show an average accuracy of 67%. The aspect and slope classifiers are trained and tested with 11449 labeled roofs in the Zurich periphery area. The results for aspect and slope classification show different accuracies depending on the classes: while some classes are well identified, other under-represented classes remain challenging to detect.

Automatic Classification of Time-variable X-Ray Sources

NASA Astrophysics Data System (ADS)

Lo, Kitty K.; Farrell, Sean; Murphy, Tara; Gaensler, B. M.

2014-05-01

To maximize the discovery potential of future synoptic surveys, especially in the field of transient science, it will be necessary to use automatic classification to identify some of the astronomical sources. The data mining technique of supervised classification is suitable for this problem. Here, we present a supervised learning method to automatically classify variable X-ray sources in the Second XMM-Newton Serendipitous Source Catalog (2XMMi-DR2). Random Forest is our classifier of choice since it is one of the most accurate learning algorithms available. Our training set consists of 873 variable sources and their features are derived from time series, spectra, and other multi-wavelength contextual information. The 10 fold cross validation accuracy of the training data is ~97% on a 7 class data set. We applied the trained classification model to 411 unknown variable 2XMM sources to produce a probabilistically classified catalog. Using the classification margin and the Random Forest derived outlier measure, we identified 12 anomalous sources, of which 2XMM J180658.7-500250 appears to be the most unusual source in the sample. Its X-ray spectra is suggestive of a ultraluminous X-ray source but its variability makes it highly unusual. Machine-learned classification and anomaly detection will facilitate scientific discoveries in the era of all-sky surveys.

Predicting Ascospore Release of Monilinia vaccinii-corymbosi of Blueberry with Machine Learning.

PubMed

Harteveld, Dalphy O C; Grant, Michael R; Pscheidt, Jay W; Peever, Tobin L

2017-11-01

Mummy berry, caused by Monilinia vaccinii-corymbosi, causes economic losses of highbush blueberry in the U.S. Pacific Northwest (PNW). Apothecia develop from mummified berries overwintering on soil surfaces and produce ascospores that infect tissue emerging from floral and vegetative buds. Disease control currently relies on fungicides applied on a calendar basis rather than inoculum availability. To establish a prediction model for ascospore release, apothecial development was tracked in three fields, one in western Oregon and two in northwestern Washington in 2015 and 2016. Air and soil temperature, precipitation, soil moisture, leaf wetness, relative humidity and solar radiation were monitored using in-field weather stations and Washington State University's AgWeatherNet stations. Four modeling approaches were compared: logistic regression, multivariate adaptive regression splines, artificial neural networks, and random forest. A supervised learning approach was used to train the models on two data sets: training (70%) and testing (30%). The importance of environmental factors was calculated for each model separately. Soil temperature, soil moisture, and solar radiation were identified as the most important factors influencing ascospore release. Random forest models, with 78% accuracy, showed the best performance compared with the other models. Results of this research helps PNW blueberry growers to optimize fungicide use and reduce production costs.

Prediction of Protein-Protein Interaction Sites by Random Forest Algorithm with mRMR and IFS

PubMed Central

Li, Bi-Qing; Feng, Kai-Yan; Chen, Lei; Huang, Tao; Cai, Yu-Dong

2012-01-01

Prediction of protein-protein interaction (PPI) sites is one of the most challenging problems in computational biology. Although great progress has been made by employing various machine learning approaches with numerous characteristic features, the problem is still far from being solved. In this study, we developed a novel predictor based on Random Forest (RF) algorithm with the Minimum Redundancy Maximal Relevance (mRMR) method followed by incremental feature selection (IFS). We incorporated features of physicochemical/biochemical properties, sequence conservation, residual disorder, secondary structure and solvent accessibility. We also included five 3D structural features to predict protein-protein interaction sites and achieved an overall accuracy of 0.672997 and MCC of 0.347977. Feature analysis showed that 3D structural features such as Depth Index (DPX) and surface curvature (SC) contributed most to the prediction of protein-protein interaction sites. It was also shown via site-specific feature analysis that the features of individual residues from PPI sites contribute most to the determination of protein-protein interaction sites. It is anticipated that our prediction method will become a useful tool for identifying PPI sites, and that the feature analysis described in this paper will provide useful insights into the mechanisms of interaction. PMID:22937126

Analyzing Body Movements within the Laban Effort Framework Using a Single Accelerometer

PubMed Central

Kikhia, Basel; Gomez, Miguel; Jiménez, Lara Lorna; Hallberg, Josef; Karvonen, Niklas; Synnes, Kåre

2014-01-01

This article presents a study on analyzing body movements by using a single accelerometer sensor. The investigated categories of body movements belong to the Laban Effort Framework: Strong—Light, Free—Bound and Sudden—Sustained. All body movements were represented by a set of activities used for data collection. The calculated accuracy of detecting the body movements was based on collecting data from a single wireless tri-axial accelerometer sensor. Ten healthy subjects collected data from three body locations (chest, wrist and thigh) simultaneously in order to analyze the locations comparatively. The data was then processed and analyzed using Machine Learning techniques. The wrist placement was found to be the best single location to record data for detecting Strong—Light body movements using the Random Forest classifier. The wrist placement was also the best location for classifying Bound—Free body movements using the SVM classifier. However, the data collected from the chest placement yielded the best results for detecting Sudden—Sustained body movements using the Random Forest classifier. The study shows that the choice of the accelerometer placement should depend on the targeted type of movement. In addition, the choice of the classifier when processing data should also depend on the chosen location and the target movement. PMID:24662408

Ensemble of trees approaches to risk adjustment for evaluating a hospital's performance.

PubMed

Liu, Yang; Traskin, Mikhail; Lorch, Scott A; George, Edward I; Small, Dylan

2015-03-01

A commonly used method for evaluating a hospital's performance on an outcome is to compare the hospital's observed outcome rate to the hospital's expected outcome rate given its patient (case) mix and service. The process of calculating the hospital's expected outcome rate given its patient mix and service is called risk adjustment (Iezzoni 1997). Risk adjustment is critical for accurately evaluating and comparing hospitals' performances since we would not want to unfairly penalize a hospital just because it treats sicker patients. The key to risk adjustment is accurately estimating the probability of an Outcome given patient characteristics. For cases with binary outcomes, the method that is commonly used in risk adjustment is logistic regression. In this paper, we consider ensemble of trees methods as alternatives for risk adjustment, including random forests and Bayesian additive regression trees (BART). Both random forests and BART are modern machine learning methods that have been shown recently to have excellent performance for prediction of outcomes in many settings. We apply these methods to carry out risk adjustment for the performance of neonatal intensive care units (NICU). We show that these ensemble of trees methods outperform logistic regression in predicting mortality among babies treated in NICU, and provide a superior method of risk adjustment compared to logistic regression.

Quantitative prediction of oral cancer risk in patients with oral leukoplakia.

PubMed

Liu, Yao; Li, Yicheng; Fu, Yue; Liu, Tong; Liu, Xiaoyong; Zhang, Xinyan; Fu, Jie; Guan, Xiaobing; Chen, Tong; Chen, Xiaoxin; Sun, Zheng

2017-07-11

Exfoliative cytology has been widely used for early diagnosis of oral squamous cell carcinoma. We have developed an oral cancer risk index using DNA index value to quantitatively assess cancer risk in patients with oral leukoplakia, but with limited success. In order to improve the performance of the risk index, we collected exfoliative cytology, histopathology, and clinical follow-up data from two independent cohorts of normal, leukoplakia and cancer subjects (training set and validation set). Peaks were defined on the basis of first derivatives with positives, and modern machine learning techniques were utilized to build statistical prediction models on the reconstructed data. Random forest was found to be the best model with high sensitivity (100%) and specificity (99.2%). Using the Peaks-Random Forest model, we constructed an index (OCRI2) as a quantitative measurement of cancer risk. Among 11 leukoplakia patients with an OCRI2 over 0.5, 4 (36.4%) developed cancer during follow-up (23 ± 20 months), whereas 3 (5.3%) of 57 leukoplakia patients with an OCRI2 less than 0.5 developed cancer (32 ± 31 months). OCRI2 is better than other methods in predicting oral squamous cell carcinoma during follow-up. In conclusion, we have developed an exfoliative cytology-based method for quantitative prediction of cancer risk in patients with oral leukoplakia.

Machine-Learning Algorithms to Automate Morphological and Functional Assessments in 2D Echocardiography.

PubMed

Narula, Sukrit; Shameer, Khader; Salem Omar, Alaa Mabrouk; Dudley, Joel T; Sengupta, Partho P

2016-11-29

Machine-learning models may aid cardiac phenotypic recognition by using features of cardiac tissue deformation. This study investigated the diagnostic value of a machine-learning framework that incorporates speckle-tracking echocardiographic data for automated discrimination of hypertrophic cardiomyopathy (HCM) from physiological hypertrophy seen in athletes (ATH). Expert-annotated speckle-tracking echocardiographic datasets obtained from 77 ATH and 62 HCM patients were used for developing an automated system. An ensemble machine-learning model with 3 different machine-learning algorithms (support vector machines, random forests, and artificial neural networks) was developed and a majority voting method was used for conclusive predictions with further K-fold cross-validation. Feature selection using an information gain (IG) algorithm revealed that volume was the best predictor for differentiating between HCM ands. ATH (IG = 0.24) followed by mid-left ventricular segmental (IG = 0.134) and average longitudinal strain (IG = 0.131). The ensemble machine-learning model showed increased sensitivity and specificity compared with early-to-late diastolic transmitral velocity ratio (p < 0.01), average early diastolic tissue velocity (e') (p < 0.01), and strain (p = 0.04). Because ATH were younger, adjusted analysis was undertaken in younger HCM patients and compared with ATH with left ventricular wall thickness >13 mm. In this subgroup analysis, the automated model continued to show equal sensitivity, but increased specificity relative to early-to-late diastolic transmitral velocity ratio, e', and strain. Our results suggested that machine-learning algorithms can assist in the discrimination of physiological versus pathological patterns of hypertrophic remodeling. This effort represents a step toward the development of a real-time, machine-learning-based system for automated interpretation of echocardiographic images, which may help novice readers with limited experience. Copyright © 2016 American College of Cardiology Foundation. Published by Elsevier Inc. All rights reserved.

Using random forest for reliable classification and cost-sensitive learning for medical diagnosis.

PubMed

Yang, Fan; Wang, Hua-zhen; Mi, Hong; Lin, Cheng-de; Cai, Wei-wen

2009-01-30

Most machine-learning classifiers output label predictions for new instances without indicating how reliable the predictions are. The applicability of these classifiers is limited in critical domains where incorrect predictions have serious consequences, like medical diagnosis. Further, the default assumption of equal misclassification costs is most likely violated in medical diagnosis. In this paper, we present a modified random forest classifier which is incorporated into the conformal predictor scheme. A conformal predictor is a transductive learning scheme, using Kolmogorov complexity to test the randomness of a particular sample with respect to the training sets. Our method show well-calibrated property that the performance can be set prior to classification and the accurate rate is exactly equal to the predefined confidence level. Further, to address the cost sensitive problem, we extend our method to a label-conditional predictor which takes into account different costs for misclassifications in different class and allows different confidence level to be specified for each class. Intensive experiments on benchmark datasets and real world applications show the resultant classifier is well-calibrated and able to control the specific risk of different class. The method of using RF outlier measure to design a nonconformity measure benefits the resultant predictor. Further, a label-conditional classifier is developed and turn to be an alternative approach to the cost sensitive learning problem that relies on label-wise predefined confidence level. The target of minimizing the risk of misclassification is achieved by specifying the different confidence level for different class.

Use of Hundreds of Electrocardiograhpic Biomarkers for Prediction of Mortality in Post-Menopausal Women: The Women’s Health Initiative

PubMed Central

Gorodeski, Eiran Z.; Ishwaran, Hemant; Kogalur, Udaya B.; Blackstone, Eugene H.; Hsich, Eileen; Zhang, Zhu-ming; Vitolins, Mara Z.; Manson, JoAnn E.; Curb, J. David; Martin, Lisa W.; Prineas, Ronald J.; Lauer, Michael S.

2013-01-01

Background Simultaneous contribution of hundreds of electrocardiographic biomarkers to prediction of long-term mortality in post-menopausal women with clinically normal resting electrocardiograms (ECGs) is unknown. Methods and Results We analyzed ECGs and all-cause mortality in 33,144 women enrolled in Women’s Health Initiative trials, who were without baseline cardiovascular disease or cancer, and had normal ECGs by Minnesota and Novacode criteria. Four hundred and seventy seven ECG biomarkers, encompassing global and individual ECG findings, were measured using computer algorithms. During a median follow-up of 8.1 years (range for survivors 0.5–11.2 years), 1,229 women died. For analyses cohort was randomly split into derivation (n=22,096, deaths=819) and validation (n=11,048, deaths=410) subsets. ECG biomarkers, demographic, and clinical characteristics were simultaneously analyzed using both traditional Cox regression and Random Survival Forest (RSF), a novel algorithmic machine-learning approach. Regression modeling failed to converge. RSF variable selection yielded 20 variables that were independently predictive of long-term mortality, 14 of which were ECG biomarkers related to autonomic tone, atrial conduction, and ventricular depolarization and repolarization. Conclusions We identified 14 ECG biomarkers from amongst hundreds that were associated with long-term prognosis using a novel random forest variable selection methodology. These were related to autonomic tone, atrial conduction, ventricular depolarization, and ventricular repolarization. Quantitative ECG biomarkers have prognostic importance, and may be markers of subclinical disease in apparently healthy post-menopausal women. PMID:21862719

Automated source classification of new transient sources

NASA Astrophysics Data System (ADS)

Oertel, M.; Kreikenbohm, A.; Wilms, J.; DeLuca, A.

2017-10-01

The EXTraS project harvests the hitherto unexplored temporal domain information buried in the serendipitous data collected by the European Photon Imaging Camera (EPIC) onboard the ESA XMM-Newton mission since its launch. This includes a search for fast transients, missed by standard image analysis, and a search and characterization of variability in hundreds of thousands of sources. We present an automated classification scheme for new transient sources in the EXTraS project. The method is as follows: source classification features of a training sample are used to train machine learning algorithms (performed in R; randomForest (Breiman, 2001) in supervised mode) which are then tested on a sample of known source classes and used for classification.

Comparison of classification algorithms for various methods of preprocessing radar images of the MSTAR base

NASA Astrophysics Data System (ADS)

Borodinov, A. A.; Myasnikov, V. V.

2018-04-01

The present work is devoted to comparing the accuracy of the known qualification algorithms in the task of recognizing local objects on radar images for various image preprocessing methods. Preprocessing involves speckle noise filtering and normalization of the object orientation in the image by the method of image moments and by a method based on the Hough transform. In comparison, the following classification algorithms are used: Decision tree; Support vector machine, AdaBoost, Random forest. The principal component analysis is used to reduce the dimension. The research is carried out on the objects from the base of radar images MSTAR. The paper presents the results of the conducted studies.

Benchmarking protein classification algorithms via supervised cross-validation.

PubMed

Kertész-Farkas, Attila; Dhir, Somdutta; Sonego, Paolo; Pacurar, Mircea; Netoteia, Sergiu; Nijveen, Harm; Kuzniar, Arnold; Leunissen, Jack A M; Kocsor, András; Pongor, Sándor

2008-04-24

Development and testing of protein classification algorithms are hampered by the fact that the protein universe is characterized by groups vastly different in the number of members, in average protein size, similarity within group, etc. Datasets based on traditional cross-validation (k-fold, leave-one-out, etc.) may not give reliable estimates on how an algorithm will generalize to novel, distantly related subtypes of the known protein classes. Supervised cross-validation, i.e., selection of test and train sets according to the known subtypes within a database has been successfully used earlier in conjunction with the SCOP database. Our goal was to extend this principle to other databases and to design standardized benchmark datasets for protein classification. Hierarchical classification trees of protein categories provide a simple and general framework for designing supervised cross-validation strategies for protein classification. Benchmark datasets can be designed at various levels of the concept hierarchy using a simple graph-theoretic distance. A combination of supervised and random sampling was selected to construct reduced size model datasets, suitable for algorithm comparison. Over 3000 new classification tasks were added to our recently established protein classification benchmark collection that currently includes protein sequence (including protein domains and entire proteins), protein structure and reading frame DNA sequence data. We carried out an extensive evaluation based on various machine-learning algorithms such as nearest neighbor, support vector machines, artificial neural networks, random forests and logistic regression, used in conjunction with comparison algorithms, BLAST, Smith-Waterman, Needleman-Wunsch, as well as 3D comparison methods DALI and PRIDE. The resulting datasets provide lower, and in our opinion more realistic estimates of the classifier performance than do random cross-validation schemes. A combination of supervised and random sampling was used to construct model datasets, suitable for algorithm comparison.

Decentralized real-time simulation of forest machines

NASA Astrophysics Data System (ADS)

Freund, Eckhard; Adam, Frank; Hoffmann, Katharina; Rossmann, Juergen; Kraemer, Michael; Schluse, Michael

2000-10-01

To develop realistic forest machine simulators is a demanding task. A useful simulator has to provide a close- to-reality simulation of the forest environment as well as the simulation of the physics of the vehicle. Customers demand a highly realistic three dimensional forestry landscape and the realistic simulation of the complex motion of the vehicle even in rough terrain in order to be able to use the simulator for operator training under close-to- reality conditions. The realistic simulation of the vehicle, especially with the driver's seat mounted on a motion platform, greatly improves the effect of immersion into the virtual reality of a simulated forest and the achievable level of education of the driver. Thus, the connection of the real control devices of forest machines to the simulation system has to be supported, i.e. the real control devices like the joysticks or the board computer system to control the crane, the aggregate etc. Beyond, the fusion of the board computer system and the simulation system is realized by means of sensors, i.e. digital and analog signals. The decentralized system structure allows several virtual reality systems to evaluate and visualize the information of the control devices and the sensors. So, while the driver is practicing, the instructor can immerse into the same virtual forest to monitor the session from his own viewpoint. In this paper, we are describing the realized structure as well as the necessary software and hardware components and application experiences.

Effectively identifying compound-protein interactions by learning from positive and unlabeled examples.

PubMed

Cheng, Zhanzhan; Zhou, Shuigeng; Wang, Yang; Liu, Hui; Guan, Jihong; Chen, Yi-Ping Phoebe

2016-05-18

Prediction of compound-protein interactions (CPIs) is to find new compound-protein pairs where a protein is targeted by at least a compound, which is a crucial step in new drug design. Currently, a number of machine learning based methods have been developed to predict new CPIs in the literature. However, as there is not yet any publicly available set of validated negative CPIs, most existing machine learning based approaches use the unknown interactions (not validated CPIs) selected randomly as the negative examples to train classifiers for predicting new CPIs. Obviously, this is not quite reasonable and unavoidably impacts the CPI prediction performance. In this paper, we simply take the unknown CPIs as unlabeled examples, and propose a new method called PUCPI (the abbreviation of PU learning for Compound-Protein Interaction identification) that employs biased-SVM (Support Vector Machine) to predict CPIs using only positive and unlabeled examples. PU learning is a class of learning methods that leans from positive and unlabeled (PU) samples. To the best of our knowledge, this is the first work that identifies CPIs using only positive and unlabeled examples. We first collect known CPIs as positive examples and then randomly select compound-protein pairs not in the positive set as unlabeled examples. For each CPI/compound-protein pair, we extract protein domains as protein features and compound substructures as chemical features, then take the tensor product of the corresponding compound features and protein features as the feature vector of the CPI/compound-protein pair. After that, biased-SVM is employed to train classifiers on different datasets of CPIs and compound-protein pairs. Experiments over various datasets show that our method outperforms six typical classifiers, including random forest, L1- and L2-regularized logistic regression, naive Bayes, SVM and k-nearest neighbor (kNN), and three types of existing CPI prediction models. Source code, datasets and related documents of PUCPI are available at: http://admis.fudan.edu.cn/projects/pucpi.html.

Estimation of Rice Crop Yields Using Random Forests in Taiwan

NASA Astrophysics Data System (ADS)

Chen, C. F.; Lin, H. S.; Nguyen, S. T.; Chen, C. R.

2017-12-01

Rice is globally one of the most important food crops, directly feeding more people than any other crops. Rice is not only the most important commodity, but also plays a critical role in the economy of Taiwan because it provides employment and income for large rural populations. The rice harvested area and production are thus monitored yearly due to the government's initiatives. Agronomic planners need such information for more precise assessment of food production to tackle issues of national food security and policymaking. This study aimed to develop a machine-learning approach using physical parameters to estimate rice crop yields in Taiwan. We processed the data for 2014 cropping seasons, following three main steps: (1) data pre-processing to construct input layers, including soil types and weather parameters (e.g., maxima and minima air temperature, precipitation, and solar radiation) obtained from meteorological stations across the country; (2) crop yield estimation using the random forests owing to its merits as it can process thousands of variables, estimate missing data, maintain the accuracy level when a large proportion of the data is missing, overcome most of over-fitting problems, and run fast and efficiently when handling large datasets; and (3) error verification. To execute the model, we separated the datasets into two groups of pixels: group-1 (70% of pixels) for training the model and group-2 (30% of pixels) for testing the model. Once the model is trained to produce small and stable out-of-bag error (i.e., the mean squared error between predicted and actual values), it can be used for estimating rice yields of cropping seasons. The results obtained from the random forests-based regression were compared with the actual yield statistics indicated the values of root mean square error (RMSE) and mean absolute error (MAE) achieved for the first rice crop were respectively 6.2% and 2.7%, while those for the second rice crop were 5.3% and 2.9%, respectively. Although there are several uncertainties attributed to the data quality of input layers, our study demonstrates the promising application of random forests for estimating rice crop yields at the national level in Taiwan. This approach could be transferable to other regions of the world for improving large-scale estimation of rice crop yields.

Probability estimation with machine learning methods for dichotomous and multicategory outcome: theory.

PubMed

Kruppa, Jochen; Liu, Yufeng; Biau, Gérard; Kohler, Michael; König, Inke R; Malley, James D; Ziegler, Andreas

2014-07-01

Probability estimation for binary and multicategory outcome using logistic and multinomial logistic regression has a long-standing tradition in biostatistics. However, biases may occur if the model is misspecified. In contrast, outcome probabilities for individuals can be estimated consistently with machine learning approaches, including k-nearest neighbors (k-NN), bagged nearest neighbors (b-NN), random forests (RF), and support vector machines (SVM). Because machine learning methods are rarely used by applied biostatisticians, the primary goal of this paper is to explain the concept of probability estimation with these methods and to summarize recent theoretical findings. Probability estimation in k-NN, b-NN, and RF can be embedded into the class of nonparametric regression learning machines; therefore, we start with the construction of nonparametric regression estimates and review results on consistency and rates of convergence. In SVMs, outcome probabilities for individuals are estimated consistently by repeatedly solving classification problems. For SVMs we review classification problem and then dichotomous probability estimation. Next we extend the algorithms for estimating probabilities using k-NN, b-NN, and RF to multicategory outcomes and discuss approaches for the multicategory probability estimation problem using SVM. In simulation studies for dichotomous and multicategory dependent variables we demonstrate the general validity of the machine learning methods and compare it with logistic regression. However, each method fails in at least one simulation scenario. We conclude with a discussion of the failures and give recommendations for selecting and tuning the methods. Applications to real data and example code are provided in a companion article (doi:10.1002/bimj.201300077). © 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.

Scalable Machine Learning for Massive Astronomical Datasets

NASA Astrophysics Data System (ADS)

Ball, Nicholas M.; Gray, A.

2014-04-01

We present the ability to perform data mining and machine learning operations on a catalog of half a billion astronomical objects. This is the result of the combination of robust, highly accurate machine learning algorithms with linear scalability that renders the applications of these algorithms to massive astronomical data tractable. We demonstrate the core algorithms kernel density estimation, K-means clustering, linear regression, nearest neighbors, random forest and gradient-boosted decision tree, singular value decomposition, support vector machine, and two-point correlation function. Each of these is relevant for astronomical applications such as finding novel astrophysical objects, characterizing artifacts in data, object classification (including for rare objects), object distances, finding the important features describing objects, density estimation of distributions, probabilistic quantities, and exploring the unknown structure of new data. The software, Skytree Server, runs on any UNIX-based machine, a virtual machine, or cloud-based and distributed systems including Hadoop. We have integrated it on the cloud computing system of the Canadian Astronomical Data Centre, the Canadian Advanced Network for Astronomical Research (CANFAR), creating the world's first cloud computing data mining system for astronomy. We demonstrate results showing the scaling of each of our major algorithms on large astronomical datasets, including the full 470,992,970 objects of the 2 Micron All-Sky Survey (2MASS) Point Source Catalog. We demonstrate the ability to find outliers in the full 2MASS dataset utilizing multiple methods, e.g., nearest neighbors. This is likely of particular interest to the radio astronomy community given, for example, that survey projects contain groups dedicated to this topic. 2MASS is used as a proof-of-concept dataset due to its convenience and availability. These results are of interest to any astronomical project with large and/or complex datasets that wishes to extract the full scientific value from its data.

Scalable Machine Learning for Massive Astronomical Datasets

NASA Astrophysics Data System (ADS)

Ball, Nicholas M.; Astronomy Data Centre, Canadian

2014-01-01

We present the ability to perform data mining and machine learning operations on a catalog of half a billion astronomical objects. This is the result of the combination of robust, highly accurate machine learning algorithms with linear scalability that renders the applications of these algorithms to massive astronomical data tractable. We demonstrate the core algorithms kernel density estimation, K-means clustering, linear regression, nearest neighbors, random forest and gradient-boosted decision tree, singular value decomposition, support vector machine, and two-point correlation function. Each of these is relevant for astronomical applications such as finding novel astrophysical objects, characterizing artifacts in data, object classification (including for rare objects), object distances, finding the important features describing objects, density estimation of distributions, probabilistic quantities, and exploring the unknown structure of new data. The software, Skytree Server, runs on any UNIX-based machine, a virtual machine, or cloud-based and distributed systems including Hadoop. We have integrated it on the cloud computing system of the Canadian Astronomical Data Centre, the Canadian Advanced Network for Astronomical Research (CANFAR), creating the world's first cloud computing data mining system for astronomy. We demonstrate results showing the scaling of each of our major algorithms on large astronomical datasets, including the full 470,992,970 objects of the 2 Micron All-Sky Survey (2MASS) Point Source Catalog. We demonstrate the ability to find outliers in the full 2MASS dataset utilizing multiple methods, e.g., nearest neighbors, and the local outlier factor. 2MASS is used as a proof-of-concept dataset due to its convenience and availability. These results are of interest to any astronomical project with large and/or complex datasets that wishes to extract the full scientific value from its data.

Machine learning-based dual-energy CT parametric mapping

NASA Astrophysics Data System (ADS)

Su, Kuan-Hao; Kuo, Jung-Wen; Jordan, David W.; Van Hedent, Steven; Klahr, Paul; Wei, Zhouping; Helo, Rose Al; Liang, Fan; Qian, Pengjiang; Pereira, Gisele C.; Rassouli, Negin; Gilkeson, Robert C.; Traughber, Bryan J.; Cheng, Chee-Wai; Muzic, Raymond F., Jr.

2018-06-01

The aim is to develop and evaluate machine learning methods for generating quantitative parametric maps of effective atomic number (Zeff), relative electron density (ρ e), mean excitation energy (I x ), and relative stopping power (RSP) from clinical dual-energy CT data. The maps could be used for material identification and radiation dose calculation. Machine learning methods of historical centroid (HC), random forest (RF), and artificial neural networks (ANN) were used to learn the relationship between dual-energy CT input data and ideal output parametric maps calculated for phantoms from the known compositions of 13 tissue substitutes. After training and model selection steps, the machine learning predictors were used to generate parametric maps from independent phantom and patient input data. Precision and accuracy were evaluated using the ideal maps. This process was repeated for a range of exposure doses, and performance was compared to that of the clinically-used dual-energy, physics-based method which served as the reference. The machine learning methods generated more accurate and precise parametric maps than those obtained using the reference method. Their performance advantage was particularly evident when using data from the lowest exposure, one-fifth of a typical clinical abdomen CT acquisition. The RF method achieved the greatest accuracy. In comparison, the ANN method was only 1% less accurate but had much better computational efficiency than RF, being able to produce parametric maps in 15 s. Machine learning methods outperformed the reference method in terms of accuracy and noise tolerance when generating parametric maps, encouraging further exploration of the techniques. Among the methods we evaluated, ANN is the most suitable for clinical use due to its combination of accuracy, excellent low-noise performance, and computational efficiency.

Machine learning-based dual-energy CT parametric mapping.

PubMed

Su, Kuan-Hao; Kuo, Jung-Wen; Jordan, David W; Van Hedent, Steven; Klahr, Paul; Wei, Zhouping; Al Helo, Rose; Liang, Fan; Qian, Pengjiang; Pereira, Gisele C; Rassouli, Negin; Gilkeson, Robert C; Traughber, Bryan J; Cheng, Chee-Wai; Muzic, Raymond F

2018-06-08

The aim is to develop and evaluate machine learning methods for generating quantitative parametric maps of effective atomic number (Z eff ), relative electron density (ρ e ), mean excitation energy (I x ), and relative stopping power (RSP) from clinical dual-energy CT data. The maps could be used for material identification and radiation dose calculation. Machine learning methods of historical centroid (HC), random forest (RF), and artificial neural networks (ANN) were used to learn the relationship between dual-energy CT input data and ideal output parametric maps calculated for phantoms from the known compositions of 13 tissue substitutes. After training and model selection steps, the machine learning predictors were used to generate parametric maps from independent phantom and patient input data. Precision and accuracy were evaluated using the ideal maps. This process was repeated for a range of exposure doses, and performance was compared to that of the clinically-used dual-energy, physics-based method which served as the reference. The machine learning methods generated more accurate and precise parametric maps than those obtained using the reference method. Their performance advantage was particularly evident when using data from the lowest exposure, one-fifth of a typical clinical abdomen CT acquisition. The RF method achieved the greatest accuracy. In comparison, the ANN method was only 1% less accurate but had much better computational efficiency than RF, being able to produce parametric maps in 15 s. Machine learning methods outperformed the reference method in terms of accuracy and noise tolerance when generating parametric maps, encouraging further exploration of the techniques. Among the methods we evaluated, ANN is the most suitable for clinical use due to its combination of accuracy, excellent low-noise performance, and computational efficiency.

Multi-temporal Land Use Mapping of Coastal Wetlands Area using Machine Learning in Google Earth Engine

NASA Astrophysics Data System (ADS)

Farda, N. M.

2017-12-01

Coastal wetlands provide ecosystem services essential to people and the environment. Changes in coastal wetlands, especially on land use, are important to monitor by utilizing multi-temporal imagery. The Google Earth Engine (GEE) provides many machine learning algorithms (10 algorithms) that are very useful for extracting land use from imagery. The research objective is to explore machine learning in Google Earth Engine and its accuracy for multi-temporal land use mapping of coastal wetland area. Landsat 3 MSS (1978), Landsat 5 TM (1991), Landsat 7 ETM+ (2001), and Landsat 8 OLI (2014) images located in Segara Anakan lagoon are selected to represent multi temporal images. The input for machine learning are visible and near infrared bands, PCA band, invers PCA bands, bare soil index, vegetation index, wetness index, elevation from ASTER GDEM, and GLCM (Harralick) texture, and also polygon samples in 140 locations. There are 10 machine learning algorithms applied to extract coastal wetlands land use from Landsat imagery. The algorithms are Fast Naive Bayes, CART (Classification and Regression Tree), Random Forests, GMO Max Entropy, Perceptron (Multi Class Perceptron), Winnow, Voting SVM, Margin SVM, Pegasos (Primal Estimated sub-GrAdient SOlver for Svm), IKPamir (Intersection Kernel Passive Aggressive Method for Information Retrieval, SVM). Machine learning in Google Earth Engine are very helpful in multi-temporal land use mapping, the highest accuracy for land use mapping of coastal wetland is CART with 96.98 % Overall Accuracy using K-Fold Cross Validation (K = 10). GEE is particularly useful for multi-temporal land use mapping with ready used image and classification algorithms, and also very challenging for other applications.

Use of Machine Learning Classifiers and Sensor Data to Detect Neurological Deficit in Stroke Patients.

PubMed

Park, Eunjeong; Chang, Hyuk-Jae; Nam, Hyo Suk

2017-04-18

The pronator drift test (PDT), a neurological examination, is widely used in clinics to measure motor weakness of stroke patients. The aim of this study was to develop a PDT tool with machine learning classifiers to detect stroke symptoms based on quantification of proximal arm weakness using inertial sensors and signal processing. We extracted features of drift and pronation from accelerometer signals of wearable devices on the inner wrists of 16 stroke patients and 10 healthy controls. Signal processing and feature selection approach were applied to discriminate PDT features used to classify stroke patients. A series of machine learning techniques, namely support vector machine (SVM), radial basis function network (RBFN), and random forest (RF), were implemented to discriminate stroke patients from controls with leave-one-out cross-validation. Signal processing by the PDT tool extracted a total of 12 PDT features from sensors. Feature selection abstracted the major attributes from the 12 PDT features to elucidate the dominant characteristics of proximal weakness of stroke patients using machine learning classification. Our proposed PDT classifiers had an area under the receiver operating characteristic curve (AUC) of .806 (SVM), .769 (RBFN), and .900 (RF) without feature selection, and feature selection improves the AUCs to .913 (SVM), .956 (RBFN), and .975 (RF), representing an average performance enhancement of 15.3%. Sensors and machine learning methods can reliably detect stroke signs and quantify proximal arm weakness. Our proposed solution will facilitate pervasive monitoring of stroke patients. ©Eunjeong Park, Hyuk-Jae Chang, Hyo Suk Nam. Originally published in the Journal of Medical Internet Research (http://www.jmir.org), 18.04.2017.

Machine learning: a useful radiological adjunct in determination of a newly diagnosed glioma's grade and IDH status.

PubMed

De Looze, Céline; Beausang, Alan; Cryan, Jane; Loftus, Teresa; Buckley, Patrick G; Farrell, Michael; Looby, Seamus; Reilly, Richard; Brett, Francesca; Kearney, Hugh

2018-05-16

Machine learning methods have been introduced as a computer aided diagnostic tool, with applications to glioma characterisation on MRI. Such an algorithmic approach may provide a useful adjunct for a rapid and accurate diagnosis of a glioma. The aim of this study is to devise a machine learning algorithm that may be used by radiologists in routine practice to aid diagnosis of both: WHO grade and IDH mutation status in de novo gliomas. To evaluate the status quo, we interrogated the accuracy of neuroradiology reports in relation to WHO grade: grade II 96.49% (95% confidence intervals [CI] 0.88, 0.99); III 36.51% (95% CI 0.24, 0.50); IV 72.9% (95% CI 0.67, 0.78). We derived five MRI parameters from the same diagnostic brain scans, in under two minutes per case, and then supplied these data to a random forest algorithm. Machine learning resulted in a high level of accuracy in prediction of tumour grade: grade II/III; area under the receiver operating characteristic curve (AUC) = 98%, sensitivity = 0.82, specificity = 0.94; grade II/IV; AUC = 100%, sensitivity = 1.0, specificity = 1.0; grade III/IV; AUC = 97%, sensitivity = 0.83, specificity = 0.97. Furthermore, machine learning also facilitated the discrimination of IDH status: AUC of 88%, sensitivity = 0.81, specificity = 0.77. These data demonstrate the ability of machine learning to accurately classify diffuse gliomas by both WHO grade and IDH status from routine MRI alone-without significant image processing, which may facilitate usage as a diagnostic adjunct in clinical practice.

A Global Study of GPP focusing on Light Use Efficiency in a Random Forest Regression Model

NASA Astrophysics Data System (ADS)

Fang, W.; Wei, S.; Yi, C.; Hendrey, G. R.

2016-12-01

Light use efficiency (LUE) is at the core of mechanistic modeling of global gross primary production (GPP). However, most LUE estimates in global models are satellite-based and coarsely measured with emphasis on environmental variables. Others are from eddy covariance towers with much greater spatial and temporal data quality and emphasis on mechanistic processes, but in a limited number of sites. In this paper, we conducted a comprehensive global study of tower-based LUE from 237 FLUXNET towers, and scaled up LUEs from in-situ tower level to global biome level. We integrated key environmental and biological variables into the tower-based LUE estimates, at 0.5o x 0.5o grid-cell resolution, using a random forest regression (RFR) approach. We then developed an RFR-LUE-GPP model using the grid-cell LUE data, and compared it to a tower-LUE-GPP model by the conventional way of treating LUE as a series of biome-specific constants. In order to calibrate the LUE models, we developed a data-driven RFR-GPP model using a random forest regression method. Our results showed that LUE varies largely with latitude. We estimated a global area-weighted average of LUE at 1.21 gC m-2 MJ-1 APAR, which led to an estimated global GPP of 102.9 Gt C /year from 2000 to 2005. The tower-LUE-GPP model tended to overestimate forest GPP in tropical and boreal regions. Large uncertainties exist in GPP estimates over sparsely vegetated areas covered by savannas and woody savannas around the middle to low latitudes (i.g. 20oS to 40oS and 5oN to 15oN) due to lack of available data. Model results were improved by incorporating Köppen climate types to represent climate /meteorological information in machine learning modeling. This shed new light on the recognized issues of climate dependence of spring onset of photosynthesis and the challenges in modeling the biome GPP of evergreen broad leaf forests (EBF) accurately. The divergent responses of GPP to temperature and precipitation at mid-high latitudes and at mid-low latitudes echoed the necessity of modeling GPP separately by latitudes. This work provided a global distribution of LUE estimate, and developed a comprehensive algorithm modeling global terrestrial carbon with high spatial and temporal resolutions.

Machine learning strategies for systems with invariance properties

NASA Astrophysics Data System (ADS)

Ling, Julia; Jones, Reese; Templeton, Jeremy

2016-08-01

In many scientific fields, empirical models are employed to facilitate computational simulations of engineering systems. For example, in fluid mechanics, empirical Reynolds stress closures enable computationally-efficient Reynolds Averaged Navier Stokes simulations. Likewise, in solid mechanics, constitutive relations between the stress and strain in a material are required in deformation analysis. Traditional methods for developing and tuning empirical models usually combine physical intuition with simple regression techniques on limited data sets. The rise of high performance computing has led to a growing availability of high fidelity simulation data. These data open up the possibility of using machine learning algorithms, such as random forests or neural networks, to develop more accurate and general empirical models. A key question when using data-driven algorithms to develop these empirical models is how domain knowledge should be incorporated into the machine learning process. This paper will specifically address physical systems that possess symmetry or invariance properties. Two different methods for teaching a machine learning model an invariance property are compared. In the first method, a basis of invariant inputs is constructed, and the machine learning model is trained upon this basis, thereby embedding the invariance into the model. In the second method, the algorithm is trained on multiple transformations of the raw input data until the model learns invariance to that transformation. Results are discussed for two case studies: one in turbulence modeling and one in crystal elasticity. It is shown that in both cases embedding the invariance property into the input features yields higher performance at significantly reduced computational training costs.

Sensitivity and specificity of machine learning classifiers for glaucoma diagnosis using Spectral Domain OCT and standard automated perimetry.

PubMed

Silva, Fabrício R; Vidotti, Vanessa G; Cremasco, Fernanda; Dias, Marcelo; Gomi, Edson S; Costa, Vital P

2013-01-01

To evaluate the sensitivity and specificity of machine learning classifiers (MLCs) for glaucoma diagnosis using Spectral Domain OCT (SD-OCT) and standard automated perimetry (SAP). Observational cross-sectional study. Sixty two glaucoma patients and 48 healthy individuals were included. All patients underwent a complete ophthalmologic examination, achromatic standard automated perimetry (SAP) and retinal nerve fiber layer (RNFL) imaging with SD-OCT (Cirrus HD-OCT; Carl Zeiss Meditec Inc., Dublin, California). Receiver operating characteristic (ROC) curves were obtained for all SD-OCT parameters and global indices of SAP. Subsequently, the following MLCs were tested using parameters from the SD-OCT and SAP: Bagging (BAG), Naive-Bayes (NB), Multilayer Perceptron (MLP), Radial Basis Function (RBF), Random Forest (RAN), Ensemble Selection (ENS), Classification Tree (CTREE), Ada Boost M1(ADA),Support Vector Machine Linear (SVML) and Support Vector Machine Gaussian (SVMG). Areas under the receiver operating characteristic curves (aROC) obtained for isolated SAP and OCT parameters were compared with MLCs using OCT+SAP data. Combining OCT and SAP data, MLCs' aROCs varied from 0.777(CTREE) to 0.946 (RAN).The best OCT+SAP aROC obtained with RAN (0.946) was significantly larger the best single OCT parameter (p<0.05), but was not significantly different from the aROC obtained with the best single SAP parameter (p=0.19). Machine learning classifiers trained on OCT and SAP data can successfully discriminate between healthy and glaucomatous eyes. The combination of OCT and SAP measurements improved the diagnostic accuracy compared with OCT data alone.

Machine learning strategies for systems with invariance properties

DOE Office of Scientific and Technical Information (OSTI.GOV)

Ling, Julia; Jones, Reese E.; Templeton, Jeremy Alan

Here, in many scientific fields, empirical models are employed to facilitate computational simulations of engineering systems. For example, in fluid mechanics, empirical Reynolds stress closures enable computationally-efficient Reynolds-Averaged Navier-Stokes simulations. Likewise, in solid mechanics, constitutive relations between the stress and strain in a material are required in deformation analysis. Traditional methods for developing and tuning empirical models usually combine physical intuition with simple regression techniques on limited data sets. The rise of high-performance computing has led to a growing availability of high-fidelity simulation data, which open up the possibility of using machine learning algorithms, such as random forests or neuralmore » networks, to develop more accurate and general empirical models. A key question when using data-driven algorithms to develop these models is how domain knowledge should be incorporated into the machine learning process. This paper will specifically address physical systems that possess symmetry or invariance properties. Two different methods for teaching a machine learning model an invariance property are compared. In the first , a basis of invariant inputs is constructed, and the machine learning model is trained upon this basis, thereby embedding the invariance into the model. In the second method, the algorithm is trained on multiple transformations of the raw input data until the model learns invariance to that transformation. Results are discussed for two case studies: one in turbulence modeling and one in crystal elasticity. It is shown that in both cases embedding the invariance property into the input features yields higher performance with significantly reduced computational training costs.« less

Machine learning strategies for systems with invariance properties

DOE PAGES

Ling, Julia; Jones, Reese E.; Templeton, Jeremy Alan

2016-05-06

Here, in many scientific fields, empirical models are employed to facilitate computational simulations of engineering systems. For example, in fluid mechanics, empirical Reynolds stress closures enable computationally-efficient Reynolds-Averaged Navier-Stokes simulations. Likewise, in solid mechanics, constitutive relations between the stress and strain in a material are required in deformation analysis. Traditional methods for developing and tuning empirical models usually combine physical intuition with simple regression techniques on limited data sets. The rise of high-performance computing has led to a growing availability of high-fidelity simulation data, which open up the possibility of using machine learning algorithms, such as random forests or neuralmore » networks, to develop more accurate and general empirical models. A key question when using data-driven algorithms to develop these models is how domain knowledge should be incorporated into the machine learning process. This paper will specifically address physical systems that possess symmetry or invariance properties. Two different methods for teaching a machine learning model an invariance property are compared. In the first , a basis of invariant inputs is constructed, and the machine learning model is trained upon this basis, thereby embedding the invariance into the model. In the second method, the algorithm is trained on multiple transformations of the raw input data until the model learns invariance to that transformation. Results are discussed for two case studies: one in turbulence modeling and one in crystal elasticity. It is shown that in both cases embedding the invariance property into the input features yields higher performance with significantly reduced computational training costs.« less

Toward Automating HIV Identification: Machine Learning for Rapid Identification of HIV-Related Social Media Data.

PubMed

Young, Sean D; Yu, Wenchao; Wang, Wei

2017-02-01

"Social big data" from technologies such as social media, wearable devices, and online searches continue to grow and can be used as tools for HIV research. Although researchers can uncover patterns and insights associated with HIV trends and transmission, the review process is time consuming and resource intensive. Machine learning methods derived from computer science might be used to assist HIV domain experts by learning how to rapidly and accurately identify patterns associated with HIV from a large set of social data. Using an existing social media data set that was associated with HIV and coded by an HIV domain expert, we tested whether 4 commonly used machine learning methods could learn the patterns associated with HIV risk behavior. We used the 10-fold cross-validation method to examine the speed and accuracy of these models in applying that knowledge to detect HIV content in social media data. Logistic regression and random forest resulted in the highest accuracy in detecting HIV-related social data (85.3%), whereas the Ridge Regression Classifier resulted in the lowest accuracy. Logistic regression yielded the fastest processing time (16.98 seconds). Machine learning can enable social big data to become a new and important tool in HIV research, helping to create a new field of "digital HIV epidemiology." If a domain expert can identify patterns in social data associated with HIV risk or HIV transmission, machine learning models could quickly and accurately learn those associations and identify potential HIV patterns in large social data sets.

Comparison of machine-learning algorithms to build a predictive model for detecting undiagnosed diabetes - ELSA-Brasil: accuracy study.

PubMed

Olivera, André Rodrigues; Roesler, Valter; Iochpe, Cirano; Schmidt, Maria Inês; Vigo, Álvaro; Barreto, Sandhi Maria; Duncan, Bruce Bartholow

2017-01-01

Type 2 diabetes is a chronic disease associated with a wide range of serious health complications that have a major impact on overall health. The aims here were to develop and validate predictive models for detecting undiagnosed diabetes using data from the Longitudinal Study of Adult Health (ELSA-Brasil) and to compare the performance of different machine-learning algorithms in this task. Comparison of machine-learning algorithms to develop predictive models using data from ELSA-Brasil. After selecting a subset of 27 candidate variables from the literature, models were built and validated in four sequential steps: (i) parameter tuning with tenfold cross-validation, repeated three times; (ii) automatic variable selection using forward selection, a wrapper strategy with four different machine-learning algorithms and tenfold cross-validation (repeated three times), to evaluate each subset of variables; (iii) error estimation of model parameters with tenfold cross-validation, repeated ten times; and (iv) generalization testing on an independent dataset. The models were created with the following machine-learning algorithms: logistic regression, artificial neural network, naïve Bayes, K-nearest neighbor and random forest. The best models were created using artificial neural networks and logistic regression. -These achieved mean areas under the curve of, respectively, 75.24% and 74.98% in the error estimation step and 74.17% and 74.41% in the generalization testing step. Most of the predictive models produced similar results, and demonstrated the feasibility of identifying individuals with highest probability of having undiagnosed diabetes, through easily-obtained clinical data.

Predicting Pre-planting Risk of Stagonospora nodorum blotch in Winter Wheat Using Machine Learning Models.

PubMed

Mehra, Lucky K; Cowger, Christina; Gross, Kevin; Ojiambo, Peter S

2016-01-01

Pre-planting factors have been associated with the late-season severity of Stagonospora nodorum blotch (SNB), caused by the fungal pathogen Parastagonospora nodorum, in winter wheat (Triticum aestivum). The relative importance of these factors in the risk of SNB has not been determined and this knowledge can facilitate disease management decisions prior to planting of the wheat crop. In this study, we examined the performance of multiple regression (MR) and three machine learning algorithms namely artificial neural networks, categorical and regression trees, and random forests (RF), in predicting the pre-planting risk of SNB in wheat. Pre-planting factors tested as potential predictor variables were cultivar resistance, latitude, longitude, previous crop, seeding rate, seed treatment, tillage type, and wheat residue. Disease severity assessed at the end of the growing season was used as the response variable. The models were developed using 431 disease cases (unique combinations of predictors) collected from 2012 to 2014 and these cases were randomly divided into training, validation, and test datasets. Models were evaluated based on the regression of observed against predicted severity values of SNB, sensitivity-specificity ROC analysis, and the Kappa statistic. A strong relationship was observed between late-season severity of SNB and specific pre-planting factors in which latitude, longitude, wheat residue, and cultivar resistance were the most important predictors. The MR model explained 33% of variability in the data, while machine learning models explained 47 to 79% of the total variability. Similarly, the MR model correctly classified 74% of the disease cases, while machine learning models correctly classified 81 to 83% of these cases. Results show that the RF algorithm, which explained 79% of the variability within the data, was the most accurate in predicting the risk of SNB, with an accuracy rate of 93%. The RF algorithm could allow early assessment of the risk of SNB, facilitating sound disease management decisions prior to planting of wheat.

A Comparative Object-Based Sugarcane Classification from Sentinel-2 Data Using Random Forests and Support Vector Machines

NASA Astrophysics Data System (ADS)

Chen, C. R.; Chen, C. F.; Nguyen, S. T.; Lau, K.; Lay, J. G.

2016-12-01

Sugarcane mostly grown in tropical and subtropical regions is one of the important commercial crops worldwide, providing significant employment, foreign exchange earnings, and other social and environmental benefits. The sugar industry is a vital component of Belize's economy as it provides employment to 15% of the country's population and 60% of the national agricultural exports. Sugarcane mapping is thus an important task due to official initiatives to provide reliable information on sugarcane-growing areas in respect to improved accuracy in monitoring sugarcane production and yield estimates. Policymakers need such monitoring information to formulate timely plans to ensure sustainably socioeconomic development. Sugarcane monitoring in Belize is traditionally carried out through time-consuming and costly field surveys. Remote sensing is an indispensable tool for crop monitoring on national, regional and global scales. The use of high and low resolution satellites for sugarcane monitoring in Belize is often restricted due to cost limitations and mixed pixel problems because sugarcane fields are small and fragmental. With the launch of Sentinel-2 satellite, it is possible to collectively map small patches of sugarcane fields over a large region as the data are free of charge and have high spectral, spatial, and temporal resolutions. This study aims to develop an object-based classification approach to comparatively map sugarcane fields in Belize from Sentinel-2 data using random forests (RF) and support vector machines (SVM). The data were processed through four main steps: (1) data pre-processing, (2) image segmentation, (3) sugarcane classification, and (4) accuracy assessment. The mapping results compared with the ground reference data indicated satisfactory results. The overall accuracies and Kappa coefficients were generally higher than 80% and 0.7, in both cases. The RF produced slightly more accurate mapping results than SVM. This study demonstrates the realization of the potential application of Sentinel-2 data for sugarcane mapping in Belize with the aid of RF and SVM methods. The methods are thus proposed for monitoring purposes in the country.

ToxiM: A Toxicity Prediction Tool for Small Molecules Developed Using Machine Learning and Chemoinformatics Approaches.

PubMed

Sharma, Ashok K; Srivastava, Gopal N; Roy, Ankita; Sharma, Vineet K

2017-01-01

The experimental methods for the prediction of molecular toxicity are tedious and time-consuming tasks. Thus, the computational approaches could be used to develop alternative methods for toxicity prediction. We have developed a tool for the prediction of molecular toxicity along with the aqueous solubility and permeability of any molecule/metabolite. Using a comprehensive and curated set of toxin molecules as a training set, the different chemical and structural based features such as descriptors and fingerprints were exploited for feature selection, optimization and development of machine learning based classification and regression models. The compositional differences in the distribution of atoms were apparent between toxins and non-toxins, and hence, the molecular features were used for the classification and regression. On 10-fold cross-validation, the descriptor-based, fingerprint-based and hybrid-based classification models showed similar accuracy (93%) and Matthews's correlation coefficient (0.84). The performances of all the three models were comparable (Matthews's correlation coefficient = 0.84-0.87) on the blind dataset. In addition, the regression-based models using descriptors as input features were also compared and evaluated on the blind dataset. Random forest based regression model for the prediction of solubility performed better ( R 2 = 0.84) than the multi-linear regression (MLR) and partial least square regression (PLSR) models, whereas, the partial least squares based regression model for the prediction of permeability (caco-2) performed better ( R 2 = 0.68) in comparison to the random forest and MLR based regression models. The performance of final classification and regression models was evaluated using the two validation datasets including the known toxins and commonly used constituents of health products, which attests to its accuracy. The ToxiM web server would be a highly useful and reliable tool for the prediction of toxicity, solubility, and permeability of small molecules.

ToxiM: A Toxicity Prediction Tool for Small Molecules Developed Using Machine Learning and Chemoinformatics Approaches

PubMed Central

Sharma, Ashok K.; Srivastava, Gopal N.; Roy, Ankita; Sharma, Vineet K.

2017-01-01

The experimental methods for the prediction of molecular toxicity are tedious and time-consuming tasks. Thus, the computational approaches could be used to develop alternative methods for toxicity prediction. We have developed a tool for the prediction of molecular toxicity along with the aqueous solubility and permeability of any molecule/metabolite. Using a comprehensive and curated set of toxin molecules as a training set, the different chemical and structural based features such as descriptors and fingerprints were exploited for feature selection, optimization and development of machine learning based classification and regression models. The compositional differences in the distribution of atoms were apparent between toxins and non-toxins, and hence, the molecular features were used for the classification and regression. On 10-fold cross-validation, the descriptor-based, fingerprint-based and hybrid-based classification models showed similar accuracy (93%) and Matthews's correlation coefficient (0.84). The performances of all the three models were comparable (Matthews's correlation coefficient = 0.84–0.87) on the blind dataset. In addition, the regression-based models using descriptors as input features were also compared and evaluated on the blind dataset. Random forest based regression model for the prediction of solubility performed better (R2 = 0.84) than the multi-linear regression (MLR) and partial least square regression (PLSR) models, whereas, the partial least squares based regression model for the prediction of permeability (caco-2) performed better (R2 = 0.68) in comparison to the random forest and MLR based regression models. The performance of final classification and regression models was evaluated using the two validation datasets including the known toxins and commonly used constituents of health products, which attests to its accuracy. The ToxiM web server would be a highly useful and reliable tool for the prediction of toxicity, solubility, and permeability of small molecules. PMID:29249969

Unsupervised domain adaptation for early detection of drought stress in hyperspectral images

NASA Astrophysics Data System (ADS)

Schmitter, P.; Steinrücken, J.; Römer, C.; Ballvora, A.; Léon, J.; Rascher, U.; Plümer, L.

2017-09-01

Hyperspectral images can be used to uncover physiological processes in plants if interpreted properly. Machine Learning methods such as Support Vector Machines (SVM) and Random Forests have been applied to estimate development of biomass and detect and predict plant diseases and drought stress. One basic requirement of machine learning implies, that training and testing is done in the same domain and the same distribution. Different genotypes, environmental conditions, illumination and sensors violate this requirement in most practical circumstances. Here, we present an approach, which enables the detection of physiological processes by transferring the prior knowledge within an existing model into a related target domain, where no label information is available. We propose a two-step transformation of the target features, which enables a direct application of an existing model. The transformation is evaluated by an objective function including additional prior knowledge about classification and physiological processes in plants. We have applied the approach to three sets of hyperspectral images, which were acquired with different plant species in different environments observed with different sensors. It is shown, that a classification model, derived on one of the sets, delivers satisfying classification results on the transformed features of the other data sets. Furthermore, in all cases early non-invasive detection of drought stress was possible.

Skipping the real world: Classification of PolSAR images without explicit feature extraction

NASA Astrophysics Data System (ADS)

Hänsch, Ronny; Hellwich, Olaf

2018-06-01

The typical processing chain for pixel-wise classification from PolSAR images starts with an optional preprocessing step (e.g. speckle reduction), continues with extracting features projecting the complex-valued data into the real domain (e.g. by polarimetric decompositions) which are then used as input for a machine-learning based classifier, and ends in an optional postprocessing (e.g. label smoothing). The extracted features are usually hand-crafted as well as preselected and represent (a somewhat arbitrary) projection from the complex to the real domain in order to fit the requirements of standard machine-learning approaches such as Support Vector Machines or Artificial Neural Networks. This paper proposes to adapt the internal node tests of Random Forests to work directly on the complex-valued PolSAR data, which makes any explicit feature extraction obsolete. This approach leads to a classification framework with a significantly decreased computation time and memory footprint since no image features have to be computed and stored beforehand. The experimental results on one fully-polarimetric and one dual-polarimetric dataset show that, despite the simpler approach, accuracy can be maintained (decreased by only less than 2 % for the fully-polarimetric dataset) or even improved (increased by roughly 9 % for the dual-polarimetric dataset).

Spectroscopic Diagnosis of Arsenic Contamination in Agricultural Soils

PubMed Central

Shi, Tiezhu; Liu, Huizeng; Chen, Yiyun; Fei, Teng; Wang, Junjie; Wu, Guofeng

2017-01-01

This study investigated the abilities of pre-processing, feature selection and machine-learning methods for the spectroscopic diagnosis of soil arsenic contamination. The spectral data were pre-processed by using Savitzky-Golay smoothing, first and second derivatives, multiplicative scatter correction, standard normal variate, and mean centering. Principle component analysis (PCA) and the RELIEF algorithm were used to extract spectral features. Machine-learning methods, including random forests (RF), artificial neural network (ANN), radial basis function- and linear function- based support vector machine (RBF- and LF-SVM) were employed for establishing diagnosis models. The model accuracies were evaluated and compared by using overall accuracies (OAs). The statistical significance of the difference between models was evaluated by using McNemar’s test (Z value). The results showed that the OAs varied with the different combinations of pre-processing, feature selection, and classification methods. Feature selection methods could improve the modeling efficiencies and diagnosis accuracies, and RELIEF often outperformed PCA. The optimal models established by RF (OA = 86%), ANN (OA = 89%), RBF- (OA = 89%) and LF-SVM (OA = 87%) had no statistical difference in diagnosis accuracies (Z < 1.96, p < 0.05). These results indicated that it was feasible to diagnose soil arsenic contamination using reflectance spectroscopy. The appropriate combination of multivariate methods was important to improve diagnosis accuracies. PMID:28471412

Modeling Gas and Gas Hydrate Accumulation in Marine Sediments Using a K-Nearest Neighbor Machine-Learning Technique

NASA Astrophysics Data System (ADS)

Wood, W. T.; Runyan, T. E.; Palmsten, M.; Dale, J.; Crawford, C.

2016-12-01

Natural Gas (primarily methane) and gas hydrate accumulations require certain bio-geochemical, as well as physical conditions, some of which are poorly sampled and/or poorly understood. We exploit recent advances in the prediction of seafloor porosity and heat flux via machine learning techniques (e.g. Random forests and Bayesian networks) to predict the occurrence of gas and subsequently gas hydrate in marine sediments. The prediction (actually guided interpolation) of key parameters we use in this study is a K-nearest neighbor technique. KNN requires only minimal pre-processing of the data and predictors, and requires minimal run-time input so the results are almost entirely data-driven. Specifically we use new estimates of sedimentation rate and sediment type, along with recently derived compaction modeling to estimate profiles of porosity and age. We combined the compaction with seafloor heat flux to estimate temperature with depth and geologic age, which, with estimates of organic carbon, and models of methanogenesis yield limits on the production of methane. Results include geospatial predictions of gas (and gas hydrate) accumulations, with quantitative estimates of uncertainty. The Generic Earth Modeling System (GEMS) we have developed to derive the machine learning estimates is modular and easily updated with new algorithms or data.

Metagenome-Wide Association Study and Machine Learning Prediction of Bulk Soil Microbiome and Crop Productivity

PubMed Central

Chang, Hao-Xun; Haudenshield, James S.; Bowen, Charles R.; Hartman, Glen L.

2017-01-01

Areas within an agricultural field in the same season often differ in crop productivity despite having the same cropping history, crop genotype, and management practices. One hypothesis is that abiotic or biotic factors in the soils differ between areas resulting in these productivity differences. In this study, bulk soil samples collected from a high and a low productivity area from within six agronomic fields in Illinois were quantified for abiotic and biotic characteristics. Extracted DNA from these bulk soil samples were shotgun sequenced. While logistic regression analyses resulted in no significant association between crop productivity and the 26 soil characteristics, principal coordinate analysis and constrained correspondence analysis showed crop productivity explained a major proportion of the taxa variance in the bulk soil microbiome. Metagenome-wide association studies (MWAS) identified more Bradyrhizodium and Gammaproteobacteria in higher productivity areas and more Actinobacteria, Ascomycota, Planctomycetales, and Streptophyta in lower productivity areas. Machine learning using a random forest method successfully predicted productivity based on the microbiome composition with the best accuracy of 0.79 at the order level. Our study showed that crop productivity differences were associated with bulk soil microbiome composition and highlighted several nitrogen utility-related taxa. We demonstrated the merit of MWAS and machine learning for the first time in a plant-microbiome study. PMID:28421041

Automatic Classification of Tremor Severity in Parkinson's Disease Using a Wearable Device.

PubMed

Jeon, Hyoseon; Lee, Woongwoo; Park, Hyeyoung; Lee, Hong Ji; Kim, Sang Kyong; Kim, Han Byul; Jeon, Beomseok; Park, Kwang Suk

2017-09-09

Although there is clinical demand for new technology that can accurately measure Parkinsonian tremors, automatic scoring of Parkinsonian tremors using machine-learning approaches has not yet been employed. This study aims to fill this gap by proposing machine-learning algorithms as a way to predict the Unified Parkinson's Disease Rating Scale (UPDRS), which are similar to how neurologists rate scores in actual clinical practice. In this study, the tremor signals of 85 patients with Parkinson's disease (PD) were measured using a wrist-watch-type wearable device consisting of an accelerometer and a gyroscope. The displacement and angle signals were calculated from the measured acceleration and angular velocity, and the acceleration, angular velocity, displacement, and angle signals were used for analysis. Nineteen features were extracted from each signal, and the pairwise correlation strategy was used to reduce the number of feature dimensions. With the selected features, a decision tree (DT), support vector machine (SVM), discriminant analysis (DA), random forest (RF), and k -nearest-neighbor ( k NN) algorithm were explored for automatic scoring of the Parkinsonian tremor severity. The performance of the employed classifiers was analyzed using accuracy, recall, and precision, and compared to other findings in similar studies. Finally, the limitations and plans for further study are discussed.

Big genomics and clinical data analytics strategies for precision cancer prognosis.

PubMed

Ow, Ghim Siong; Kuznetsov, Vladimir A

2016-11-07

The field of personalized and precise medicine in the era of big data analytics is growing rapidly. Previously, we proposed our model of patient classification termed Prognostic Signature Vector Matching (PSVM) and identified a 37 variable signature comprising 36 let-7b associated prognostic significant mRNAs and the age risk factor that stratified large high-grade serous ovarian cancer patient cohorts into three survival-significant risk groups. Here, we investigated the predictive performance of PSVM via optimization of the prognostic variable weights, which represent the relative importance of one prognostic variable over the others. In addition, we compared several multivariate prognostic models based on PSVM with classical machine learning techniques such as K-nearest-neighbor, support vector machine, random forest, neural networks and logistic regression. Our results revealed that negative log-rank p-values provides more robust weight values as opposed to the use of other quantities such as hazard ratios, fold change, or a combination of those factors. PSVM, together with the classical machine learning classifiers were combined in an ensemble (multi-test) voting system, which collectively provides a more precise and reproducible patient stratification. The use of the multi-test system approach, rather than the search for the ideal classification/prediction method, might help to address limitations of the individual classification algorithm in specific situation.

Characteristics of Mild Cognitive Impairment Using the Thai Version of the Consortium to Establish a Registry for Alzheimer's Disease Tests: A Multivariate and Machine Learning Study.

PubMed

Tunvirachaisakul, Chavit; Supasitthumrong, Thitiporn; Tangwongchai, Sookjareon; Hemrunroj, Solaphat; Chuchuen, Phenphichcha; Tawankanjanachot, Itthipol; Likitchareon, Yuthachai; Phanthumchinda, Kamman; Sriswasdi, Sira; Maes, Michael

2018-04-04

The Consortium to Establish a Registry for Alzheimer's Disease (CERAD) developed a neuropsychological battery (CERAD-NP) to screen patients with Alzheimer's dementia. Mild cognitive impairment (MCI) has received attention as a pre-dementia stage. To delineate the CERAD-NP features of MCI and their clinical utility to externally validate MCI diagnosis. The study included 60 patients with MCI, diagnosed using the Clinical Dementia Rating, and 63 normal controls. Data were analysed employing receiver operating characteristic analysis, Linear Support Vector Machine, Random Forest, Adaptive Boosting, Neural Network models, and t-distributed stochastic neighbour embedding (t-SNE). MCI patients were best discriminated from normal controls using a combination of Wordlist Recall, Wordlist Memory, and Verbal Fluency Test. Machine learning showed that the CERAD features learned from MCI patients and controls were not strongly predictive of the diagnosis (maximal cross-validation 77.2%), whilst t-SNE showed that there is a considerable overlap between MCI and controls. The most important features of the CERAD-NP differentiating MCI from normal controls indicate impairments in episodic and semantic memory and recall. While these features significantly discriminate MCI patients from normal controls, the tests are not predictive of MCI. © 2018 S. Karger AG, Basel.

Informing the Human Plasma Protein Binding of ...

EPA Pesticide Factsheets

The free fraction of a xenobiotic in plasma (Fub) is an important determinant of chemical adsorption, distribution, metabolism, elimination, and toxicity, yet experimental plasma protein binding data is scarce for environmentally relevant chemicals. The presented work explores the merit of utilizing available pharmaceutical data to predict Fub for environmentally relevant chemicals via machine learning techniques. Quantitative structure-activity relationship (QSAR) models were constructed with k nearest neighbors (kNN), support vector machines (SVM), and random forest (RF) machine learning algorithms from a training set of 1045 pharmaceuticals. The models were then evaluated with independent test sets of pharmaceuticals (200 compounds) and environmentally relevant ToxCast chemicals (406 total, in two groups of 238 and 168 compounds). The selection of a minimal feature set of 10-15 2D molecular descriptors allowed for both informative feature interpretation and practical applicability domain assessment via a bounded box of descriptor ranges and principal component analysis. The diverse pharmaceutical and environmental chemical sets exhibit similarities in terms of chemical space (99-82% overlap), as well as comparable bias and variance in constructed learning curves. All the models exhibit significant predictability with mean absolute errors (MAE) in the range of 0.10-0.18 Fub. The models performed best for highly bound chemicals (MAE 0.07-0.12), neutrals (MAE 0

Simultaneous feature selection and parameter optimisation using an artificial ant colony: case study of melting point prediction

PubMed Central

O'Boyle, Noel M; Palmer, David S; Nigsch, Florian; Mitchell, John BO

2008-01-01

Background We present a novel feature selection algorithm, Winnowing Artificial Ant Colony (WAAC), that performs simultaneous feature selection and model parameter optimisation for the development of predictive quantitative structure-property relationship (QSPR) models. The WAAC algorithm is an extension of the modified ant colony algorithm of Shen et al. (J Chem Inf Model 2005, 45: 1024–1029). We test the ability of the algorithm to develop a predictive partial least squares model for the Karthikeyan dataset (J Chem Inf Model 2005, 45: 581–590) of melting point values. We also test its ability to perform feature selection on a support vector machine model for the same dataset. Results Starting from an initial set of 203 descriptors, the WAAC algorithm selected a PLS model with 68 descriptors which has an RMSE on an external test set of 46.6°C and R2 of 0.51. The number of components chosen for the model was 49, which was close to optimal for this feature selection. The selected SVM model has 28 descriptors (cost of 5, ε of 0.21) and an RMSE of 45.1°C and R2 of 0.54. This model outperforms a kNN model (RMSE of 48.3°C, R2 of 0.47) for the same data and has similar performance to a Random Forest model (RMSE of 44.5°C, R2 of 0.55). However it is much less prone to bias at the extremes of the range of melting points as shown by the slope of the line through the residuals: -0.43 for WAAC/SVM, -0.53 for Random Forest. Conclusion With a careful choice of objective function, the WAAC algorithm can be used to optimise machine learning and regression models that suffer from overfitting. Where model parameters also need to be tuned, as is the case with support vector machine and partial least squares models, it can optimise these simultaneously. The moving probabilities used by the algorithm are easily interpreted in terms of the best and current models of the ants, and the winnowing procedure promotes the removal of irrelevant descriptors. PMID:18959785

Severe fire weather and intensive forest management increase fire severity in a multi-ownership landscape.

PubMed

Zald, Harold S J; Dunn, Christopher J

2018-04-26

Many studies have examined how fuels, topography, climate, and fire weather influence fire severity. Less is known about how different forest management practices influence fire severity in multi-owner landscapes, despite costly and controversial suppression of wildfires that do not acknowledge ownership boundaries. In 2013, the Douglas Complex burned over 19,000 ha of Oregon & California Railroad (O&C) lands in Southwestern Oregon, USA. O&C lands are composed of a checkerboard of private industrial and federal forestland (Bureau of Land Management, BLM) with contrasting management objectives, providing a unique experimental landscape to understand how different management practices influence wildfire severity. Leveraging Landsat based estimates of fire severity (Relative differenced Normalized Burn Ratio, RdNBR) and geospatial data on fire progression, weather, topography, pre-fire forest conditions, and land ownership, we asked (1) what is the relative importance of different variables driving fire severity, and (2) is intensive plantation forestry associated with higher fire severity? Using Random Forest ensemble machine learning, we found daily fire weather was the most important predictor of fire severity, followed by stand age and ownership, followed by topographic features. Estimates of pre-fire forest biomass were not an important predictor of fire severity. Adjusting for all other predictor variables in a general least squares model incorporating spatial autocorrelation, mean predicted RdNBR was higher on private industrial forests (RdNBR 521.85 ± 18.67 [mean ± SE]) vs. BLM forests (398.87 ± 18.23) with a much greater proportion of older forests. Our findings suggest intensive plantation forestry characterized by young forests and spatially homogenized fuels, rather than pre-fire biomass, were significant drivers of wildfire severity. This has implications for perceptions of wildfire risk, shared fire management responsibilities, and developing fire resilience for multiple objectives in multi-owner landscapes. © 2018 by the Ecological Society of America.

VT0005 In Action: National Forest Biomass Inventory Using Airborne Lidar Sampling

NASA Astrophysics Data System (ADS)

Saatchi, S. S.; Xu, L.; Meyer, V.; Ferraz, A.; Yang, Y.; Shapiro, A.; Bastin, J. F.

2016-12-01

Tropical countries are required to produce robust and verifiable estimates of forest carbon stocks for successful implementation of climate change mitigation. Lack of systematic national inventory data due to access, cost, and infrastructure, has impacted the capacity of most tropical countries to accurately report the GHG emissions to the international community. Here, we report on the development of the aboveground forest carbon (AGC) map of Democratic Republic of Congo (DRC) by using the VCS (Verified Carbon Standard) methodology developed by Sassan Saatchi (VT0005) using high-resolution airborne LiDAR samples. The methodology provides the distribution of the carbon stocks in aboveground live trees of more than 150 million ha of forests at 1-ha spatial resolution in DRC using more than 430, 000 ha of systematic random airborne Lidar inventory samples of forest structure. We developed a LIDAR aboveground biomass allometry using more than 100 1-ha plots across forest types and power-law model with LIDAR height metrics and average landscape scale wood density. The methodology provided estimates of forest biomass over the entire country using two approaches: 1) mean, variance, and total carbon estimates for each forest type present in DRC using inventory statistical techniques, and 2) a wall-to-wall map of the forest biomass extrapolated using satellite radar (ALOS PALSAR), surface topography from SRTM, and spectral information from Landsat (TM) and machine learning algorithms. We present the methodology, the estimates of carbon stocks and the spatial uncertainty over the entire country. AcknowledgementsThe theoretical research was carried out partially at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration, and the design and implementation in the Democratic Republic of Congo was carried out at the Institute of Environment and Sustainability at University of California Los Angeles through the support of the International Climate Initiative of the German Ministry of Environment, Conservation and Nuclear Security, and the KFW Development Bank.

Tropical land use land cover mapping in Pará (Brazil) using discriminative Markov random fields and multi-temporal TerraSAR-X data

NASA Astrophysics Data System (ADS)

Hagensieker, Ron; Roscher, Ribana; Rosentreter, Johannes; Jakimow, Benjamin; Waske, Björn

2017-12-01

Remote sensing satellite data offer the unique possibility to map land use land cover transformations by providing spatially explicit information. However, detection of short-term processes and land use patterns of high spatial-temporal variability is a challenging task. We present a novel framework using multi-temporal TerraSAR-X data and machine learning techniques, namely discriminative Markov random fields with spatio-temporal priors, and import vector machines, in order to advance the mapping of land cover characterized by short-term changes. Our study region covers a current deforestation frontier in the Brazilian state Pará with land cover dominated by primary forests, different types of pasture land and secondary vegetation, and land use dominated by short-term processes such as slash-and-burn activities. The data set comprises multi-temporal TerraSAR-X imagery acquired over the course of the 2014 dry season, as well as optical data (RapidEye, Landsat) for reference. Results show that land use land cover is reliably mapped, resulting in spatially adjusted overall accuracies of up to 79% in a five class setting, yet limitations for the differentiation of different pasture types remain. The proposed method is applicable on multi-temporal data sets, and constitutes a feasible approach to map land use land cover in regions that are affected by high-frequent temporal changes.

Quantification of the heterogeneity of prognostic cellular biomarkers in ewing sarcoma using automated image and random survival forest analysis.

PubMed

Bühnemann, Claudia; Li, Simon; Yu, Haiyue; Branford White, Harriet; Schäfer, Karl L; Llombart-Bosch, Antonio; Machado, Isidro; Picci, Piero; Hogendoorn, Pancras C W; Athanasou, Nicholas A; Noble, J Alison; Hassan, A Bassim

2014-01-01

Driven by genomic somatic variation, tumour tissues are typically heterogeneous, yet unbiased quantitative methods are rarely used to analyse heterogeneity at the protein level. Motivated by this problem, we developed automated image segmentation of images of multiple biomarkers in Ewing sarcoma to generate distributions of biomarkers between and within tumour cells. We further integrate high dimensional data with patient clinical outcomes utilising random survival forest (RSF) machine learning. Using material from cohorts of genetically diagnosed Ewing sarcoma with EWSR1 chromosomal translocations, confocal images of tissue microarrays were segmented with level sets and watershed algorithms. Each cell nucleus and cytoplasm were identified in relation to DAPI and CD99, respectively, and protein biomarkers (e.g. Ki67, pS6, Foxo3a, EGR1, MAPK) localised relative to nuclear and cytoplasmic regions of each cell in order to generate image feature distributions. The image distribution features were analysed with RSF in relation to known overall patient survival from three separate cohorts (185 informative cases). Variation in pre-analytical processing resulted in elimination of a high number of non-informative images that had poor DAPI localisation or biomarker preservation (67 cases, 36%). The distribution of image features for biomarkers in the remaining high quality material (118 cases, 104 features per case) were analysed by RSF with feature selection, and performance assessed using internal cross-validation, rather than a separate validation cohort. A prognostic classifier for Ewing sarcoma with low cross-validation error rates (0.36) was comprised of multiple features, including the Ki67 proliferative marker and a sub-population of cells with low cytoplasmic/nuclear ratio of CD99. Through elimination of bias, the evaluation of high-dimensionality biomarker distribution within cell populations of a tumour using random forest analysis in quality controlled tumour material could be achieved. Such an automated and integrated methodology has potential application in the identification of prognostic classifiers based on tumour cell heterogeneity.

CURE-SMOTE algorithm and hybrid algorithm for feature selection and parameter optimization based on random forests.

PubMed

Ma, Li; Fan, Suohai

2017-03-14

The random forests algorithm is a type of classifier with prominent universality, a wide application range, and robustness for avoiding overfitting. But there are still some drawbacks to random forests. Therefore, to improve the performance of random forests, this paper seeks to improve imbalanced data processing, feature selection and parameter optimization. We propose the CURE-SMOTE algorithm for the imbalanced data classification problem. Experiments on imbalanced UCI data reveal that the combination of Clustering Using Representatives (CURE) enhances the original synthetic minority oversampling technique (SMOTE) algorithms effectively compared with the classification results on the original data using random sampling, Borderline-SMOTE1, safe-level SMOTE, C-SMOTE, and k-means-SMOTE. Additionally, the hybrid RF (random forests) algorithm has been proposed for feature selection and parameter optimization, which uses the minimum out of bag (OOB) data error as its objective function. Simulation results on binary and higher-dimensional data indicate that the proposed hybrid RF algorithms, hybrid genetic-random forests algorithm, hybrid particle swarm-random forests algorithm and hybrid fish swarm-random forests algorithm can achieve the minimum OOB error and show the best generalization ability. The training set produced from the proposed CURE-SMOTE algorithm is closer to the original data distribution because it contains minimal noise. Thus, better classification results are produced from this feasible and effective algorithm. Moreover, the hybrid algorithm's F-value, G-mean, AUC and OOB scores demonstrate that they surpass the performance of the original RF algorithm. Hence, this hybrid algorithm provides a new way to perform feature selection and parameter optimization.

Object-based habitat mapping using very high spatial resolution multispectral and hyperspectral imagery with LiDAR data

NASA Astrophysics Data System (ADS)

Onojeghuo, Alex Okiemute; Onojeghuo, Ajoke Ruth

2017-07-01

This study investigated the combined use of multispectral/hyperspectral imagery and LiDAR data for habitat mapping across parts of south Cumbria, North West England. The methodology adopted in this study integrated spectral information contained in pansharp QuickBird multispectral/AISA Eagle hyperspectral imagery and LiDAR-derived measures with object-based machine learning classifiers and ensemble analysis techniques. Using the LiDAR point cloud data, elevation models (such as the Digital Surface Model and Digital Terrain Model raster) and intensity features were extracted directly. The LiDAR-derived measures exploited in this study included Canopy Height Model, intensity and topographic information (i.e. mean, maximum and standard deviation). These three LiDAR measures were combined with spectral information contained in the pansharp QuickBird and Eagle MNF transformed imagery for image classification experiments. A fusion of pansharp QuickBird multispectral and Eagle MNF hyperspectral imagery with all LiDAR-derived measures generated the best classification accuracies, 89.8 and 92.6% respectively. These results were generated with the Support Vector Machine and Random Forest machine learning algorithms respectively. The ensemble analysis of all three learning machine classifiers for the pansharp QuickBird and Eagle MNF fused data outputs did not significantly increase the overall classification accuracy. Results of the study demonstrate the potential of combining either very high spatial resolution multispectral or hyperspectral imagery with LiDAR data for habitat mapping.

Integrating Natural Language Processing and Machine Learning Algorithms to Categorize Oncologic Response in Radiology Reports.

PubMed

Chen, Po-Hao; Zafar, Hanna; Galperin-Aizenberg, Maya; Cook, Tessa

2018-04-01

A significant volume of medical data remains unstructured. Natural language processing (NLP) and machine learning (ML) techniques have shown to successfully extract insights from radiology reports. However, the codependent effects of NLP and ML in this context have not been well-studied. Between April 1, 2015 and November 1, 2016, 9418 cross-sectional abdomen/pelvis CT and MR examinations containing our internal structured reporting element for cancer were separated into four categories: Progression, Stable Disease, Improvement, or No Cancer. We combined each of three NLP techniques with five ML algorithms to predict the assigned label using the unstructured report text and compared the performance of each combination. The three NLP algorithms included term frequency-inverse document frequency (TF-IDF), term frequency weighting (TF), and 16-bit feature hashing. The ML algorithms included logistic regression (LR), random decision forest (RDF), one-vs-all support vector machine (SVM), one-vs-all Bayes point machine (BPM), and fully connected neural network (NN). The best-performing NLP model consisted of tokenized unigrams and bigrams with TF-IDF. Increasing N-gram length yielded little to no added benefit for most ML algorithms. With all parameters optimized, SVM had the best performance on the test dataset, with 90.6 average accuracy and F score of 0.813. The interplay between ML and NLP algorithms and their effect on interpretation accuracy is complex. The best accuracy is achieved when both algorithms are optimized concurrently.

Machine learning search for variable stars

NASA Astrophysics Data System (ADS)

Pashchenko, Ilya N.; Sokolovsky, Kirill V.; Gavras, Panagiotis

2018-04-01

Photometric variability detection is often considered as a hypothesis testing problem: an object is variable if the null hypothesis that its brightness is constant can be ruled out given the measurements and their uncertainties. The practical applicability of this approach is limited by uncorrected systematic errors. We propose a new variability detection technique sensitive to a wide range of variability types while being robust to outliers and underestimated measurement uncertainties. We consider variability detection as a classification problem that can be approached with machine learning. Logistic Regression (LR), Support Vector Machines (SVM), k Nearest Neighbours (kNN), Neural Nets (NN), Random Forests (RF), and Stochastic Gradient Boosting classifier (SGB) are applied to 18 features (variability indices) quantifying scatter and/or correlation between points in a light curve. We use a subset of Optical Gravitational Lensing Experiment phase two (OGLE-II) Large Magellanic Cloud (LMC) photometry (30 265 light curves) that was searched for variability using traditional methods (168 known variable objects) as the training set and then apply the NN to a new test set of 31 798 OGLE-II LMC light curves. Among 205 candidates selected in the test set, 178 are real variables, while 13 low-amplitude variables are new discoveries. The machine learning classifiers considered are found to be more efficient (select more variables and fewer false candidates) compared to traditional techniques using individual variability indices or their linear combination. The NN, SGB, SVM, and RF show a higher efficiency compared to LR and kNN.

Integrating machine learning techniques and high-resolution imagery to generate GIS-ready information for urban water consumption studies

NASA Astrophysics Data System (ADS)

Wolf, Nils; Hof, Angela

2012-10-01

Urban sprawl driven by shifts in tourism development produces new suburban landscapes of water consumption on Mediterranean coasts. Golf courses, ornamental, 'Atlantic' gardens and swimming pools are the most striking artefacts of this transformation, threatening the local water supply systems and exacerbating water scarcity. In the face of climate change, urban landscape irrigation is becoming increasingly important from a resource management point of view. This paper adopts urban remote sensing towards a targeted mapping approach using machine learning techniques and highresolution satellite imagery (WorldView-2) to generate GIS-ready information for urban water consumption studies. Swimming pools, vegetation and - as a subgroup of vegetation - turf grass are extracted as important determinants of water consumption. For image analysis, the complex nature of urban environments suggests spatial-spectral classification, i.e. the complementary use of the spectral signature and spatial descriptors. Multiscale image segmentation provides means to extract the spatial descriptors - namely object feature layers - which can be concatenated at pixel level to the spectral signature. This study assesses the value of object features using different machine learning techniques and amounts of labeled information for learning. The results indicate the benefit of the spatial-spectral approach if combined with appropriate classifiers like tree-based ensembles or support vector machines, which can handle high dimensionality. Finally, a Random Forest classifier was chosen to deliver the classified input data for the estimation of evaporative water loss and net landscape irrigation requirements.

Machine Learning for Flood Prediction in Google Earth Engine

NASA Astrophysics Data System (ADS)

Kuhn, C.; Tellman, B.; Max, S. A.; Schwarz, B.

2015-12-01

With the increasing availability of high-resolution satellite imagery, dynamic flood mapping in near real time is becoming a reachable goal for decision-makers. This talk describes a newly developed framework for predicting biophysical flood vulnerability using public data, cloud computing and machine learning. Our objective is to define an approach to flood inundation modeling using statistical learning methods deployed in a cloud-based computing platform. Traditionally, static flood extent maps grounded in physically based hydrologic models can require hours of human expertise to construct at significant financial cost. In addition, desktop modeling software and limited local server storage can impose restraints on the size and resolution of input datasets. Data-driven, cloud-based processing holds promise for predictive watershed modeling at a wide range of spatio-temporal scales. However, these benefits come with constraints. In particular, parallel computing limits a modeler's ability to simulate the flow of water across a landscape, rendering traditional routing algorithms unusable in this platform. Our project pushes these limits by testing the performance of two machine learning algorithms, Support Vector Machine (SVM) and Random Forests, at predicting flood extent. Constructed in Google Earth Engine, the model mines a suite of publicly available satellite imagery layers to use as algorithm inputs. Results are cross-validated using MODIS-based flood maps created using the Dartmouth Flood Observatory detection algorithm. Model uncertainty highlights the difficulty of deploying unbalanced training data sets based on rare extreme events.

Evaluation of machine learning algorithms for prediction of regions of high Reynolds averaged Navier Stokes uncertainty

DOE PAGES

Ling, Julia; Templeton, Jeremy Alan

2015-08-04

Reynolds Averaged Navier Stokes (RANS) models are widely used in industry to predict fluid flows, despite their acknowledged deficiencies. Not only do RANS models often produce inaccurate flow predictions, but there are very limited diagnostics available to assess RANS accuracy for a given flow configuration. If experimental or higher fidelity simulation results are not available for RANS validation, there is no reliable method to evaluate RANS accuracy. This paper explores the potential of utilizing machine learning algorithms to identify regions of high RANS uncertainty. Three different machine learning algorithms were evaluated: support vector machines, Adaboost decision trees, and random forests.more » The algorithms were trained on a database of canonical flow configurations for which validated direct numerical simulation or large eddy simulation results were available, and were used to classify RANS results on a point-by-point basis as having either high or low uncertainty, based on the breakdown of specific RANS modeling assumptions. Classifiers were developed for three different basic RANS eddy viscosity model assumptions: the isotropy of the eddy viscosity, the linearity of the Boussinesq hypothesis, and the non-negativity of the eddy viscosity. It is shown that these classifiers are able to generalize to flows substantially different from those on which they were trained. As a result, feature selection techniques, model evaluation, and extrapolation detection are discussed in the context of turbulence modeling applications.« less

Prediction of Nursing Workload in Hospital.

PubMed

Fiebig, Madlen; Hunstein, Dirk; Bartholomeyczik, Sabine

2018-01-01

A dissertation project at the Witten/Herdecke University [1] is investigating which (nursing sensitive) patient characteristics are suitable for predicting a higher or lower degree of nursing workload. For this research project four predictive modelling methods were selected. In a first step, SUPPORT VECTOR MACHINE, RANDOM FOREST, and GRADIENT BOOSTING were used to identify potential predictors from the nursing sensitive patient characteristics. The results were compared via FEATURE IMPORTANCE. To predict nursing workload the predictors identified in step 1 were modelled using MULTINOMIAL LOGISTIC REGRESSION. First results from the data mining process will be presented. A prognostic determination of nursing workload can be used not only as a basis for human resource planning in hospital, but also to respond to health policy issues.

Identification of compound-protein interactions through the analysis of gene ontology, KEGG enrichment for proteins and molecular fragments of compounds.

PubMed

Chen, Lei; Zhang, Yu-Hang; Zheng, Mingyue; Huang, Tao; Cai, Yu-Dong

2016-12-01

Compound-protein interactions play important roles in every cell via the recognition and regulation of specific functional proteins. The correct identification of compound-protein interactions can lead to a good comprehension of this complicated system and provide useful input for the investigation of various attributes of compounds and proteins. In this study, we attempted to understand this system by extracting properties from both proteins and compounds, in which proteins were represented by gene ontology and KEGG pathway enrichment scores and compounds were represented by molecular fragments. Advanced feature selection methods, including minimum redundancy maximum relevance, incremental feature selection, and the basic machine learning algorithm random forest, were used to analyze these properties and extract core factors for the determination of actual compound-protein interactions. Compound-protein interactions reported in The Binding Databases were used as positive samples. To improve the reliability of the results, the analytic procedure was executed five times using different negative samples. Simultaneously, five optimal prediction methods based on a random forest and yielding maximum MCCs of approximately 77.55 % were constructed and may be useful tools for the prediction of compound-protein interactions. This work provides new clues to understanding the system of compound-protein interactions by analyzing extracted core features. Our results indicate that compound-protein interactions are related to biological processes involving immune, developmental and hormone-associated pathways.

Floodplain Mapping for the Continental United States Using Machine Learning Techniques and Watershed Characteristics

NASA Astrophysics Data System (ADS)

Jafarzadegan, K.; Merwade, V.; Saksena, S.

2017-12-01

Using conventional hydrodynamic methods for floodplain mapping in large-scale and data-scarce regions is problematic due to the high cost of these methods, lack of reliable data and uncertainty propagation. In this study a new framework is proposed to generate 100-year floodplains for any gauged or ungauged watershed across the United States (U.S.). This framework uses Flood Insurance Rate Maps (FIRMs), topographic, climatic and land use data which are freely available for entire U.S. for floodplain mapping. The framework consists of three components, including a Random Forest classifier for watershed classification, a Probabilistic Threshold Binary Classifier (PTBC) for generating the floodplains, and a lookup table for linking the Random Forest classifier to the PTBC. The effectiveness and reliability of the proposed framework is tested on 145 watersheds from various geographical locations in the U.S. The validation results show that around 80 percent of total watersheds are predicted well, 14 percent have acceptable fit and less than five percent are predicted poorly compared to FIRMs. Another advantage of this framework is its ability in generating floodplains for all small rivers and tributaries. Due to the high accuracy and efficiency of this framework, it can be used as a preliminary decision making tool to generate 100-year floodplain maps for data-scarce regions and all tributaries where hydrodynamic methods are difficult to use.

Using random forest for the risk assessment of coal-floor water inrush in Panjiayao Coal Mine, northern China

NASA Astrophysics Data System (ADS)

Zhao, Dekang; Wu, Qiang; Cui, Fangpeng; Xu, Hua; Zeng, Yifan; Cao, Yufei; Du, Yuanze

2018-04-01

Coal-floor water-inrush incidents account for a large proportion of coal mine disasters in northern China, and accurate risk assessment is crucial for safe coal production. A novel and promising assessment model for water inrush is proposed based on random forest (RF), which is a powerful intelligent machine-learning algorithm. RF has considerable advantages, including high classification accuracy and the capability to evaluate the importance of variables; in particularly, it is robust in dealing with the complicated and non-linear problems inherent in risk assessment. In this study, the proposed model is applied to Panjiayao Coal Mine, northern China. Eight factors were selected as evaluation indices according to systematic analysis of the geological conditions and a field survey of the study area. Risk assessment maps were generated based on RF, and the probabilistic neural network (PNN) model was also used for risk assessment as a comparison. The results demonstrate that the two methods are consistent in the risk assessment of water inrush at the mine, and RF shows a better performance compared to PNN with an overall accuracy higher by 6.67%. It is concluded that RF is more practicable to assess the water-inrush risk than PNN. The presented method will be helpful in avoiding water inrush and also can be extended to various engineering applications.

Automatic classification of time-variable X-ray sources

DOE Office of Scientific and Technical Information (OSTI.GOV)

Lo, Kitty K.; Farrell, Sean; Murphy, Tara

2014-05-01

To maximize the discovery potential of future synoptic surveys, especially in the field of transient science, it will be necessary to use automatic classification to identify some of the astronomical sources. The data mining technique of supervised classification is suitable for this problem. Here, we present a supervised learning method to automatically classify variable X-ray sources in the Second XMM-Newton Serendipitous Source Catalog (2XMMi-DR2). Random Forest is our classifier of choice since it is one of the most accurate learning algorithms available. Our training set consists of 873 variable sources and their features are derived from time series, spectra, andmore » other multi-wavelength contextual information. The 10 fold cross validation accuracy of the training data is ∼97% on a 7 class data set. We applied the trained classification model to 411 unknown variable 2XMM sources to produce a probabilistically classified catalog. Using the classification margin and the Random Forest derived outlier measure, we identified 12 anomalous sources, of which 2XMM J180658.7–500250 appears to be the most unusual source in the sample. Its X-ray spectra is suggestive of a ultraluminous X-ray source but its variability makes it highly unusual. Machine-learned classification and anomaly detection will facilitate scientific discoveries in the era of all-sky surveys.« less

Global patterns and predictions of seafloor biomass using random forests.

PubMed

Wei, Chih-Lin; Rowe, Gilbert T; Escobar-Briones, Elva; Boetius, Antje; Soltwedel, Thomas; Caley, M Julian; Soliman, Yousria; Huettmann, Falk; Qu, Fangyuan; Yu, Zishan; Pitcher, C Roland; Haedrich, Richard L; Wicksten, Mary K; Rex, Michael A; Baguley, Jeffrey G; Sharma, Jyotsna; Danovaro, Roberto; MacDonald, Ian R; Nunnally, Clifton C; Deming, Jody W; Montagna, Paul; Lévesque, Mélanie; Weslawski, Jan Marcin; Wlodarska-Kowalczuk, Maria; Ingole, Baban S; Bett, Brian J; Billett, David S M; Yool, Andrew; Bluhm, Bodil A; Iken, Katrin; Narayanaswamy, Bhavani E

2010-12-30

A comprehensive seafloor biomass and abundance database has been constructed from 24 oceanographic institutions worldwide within the Census of Marine Life (CoML) field projects. The machine-learning algorithm, Random Forests, was employed to model and predict seafloor standing stocks from surface primary production, water-column integrated and export particulate organic matter (POM), seafloor relief, and bottom water properties. The predictive models explain 63% to 88% of stock variance among the major size groups. Individual and composite maps of predicted global seafloor biomass and abundance are generated for bacteria, meiofauna, macrofauna, and megafauna (invertebrates and fishes). Patterns of benthic standing stocks were positive functions of surface primary production and delivery of the particulate organic carbon (POC) flux to the seafloor. At a regional scale, the census maps illustrate that integrated biomass is highest at the poles, on continental margins associated with coastal upwelling and with broad zones associated with equatorial divergence. Lowest values are consistently encountered on the central abyssal plains of major ocean basins The shift of biomass dominance groups with depth is shown to be affected by the decrease in average body size rather than abundance, presumably due to decrease in quantity and quality of food supply. This biomass census and associated maps are vital components of mechanistic deep-sea food web models and global carbon cycling, and as such provide fundamental information that can be incorporated into evidence-based management.

Global Patterns and Predictions of Seafloor Biomass Using Random Forests

PubMed Central

Wei, Chih-Lin; Rowe, Gilbert T.; Escobar-Briones, Elva; Boetius, Antje; Soltwedel, Thomas; Caley, M. Julian; Soliman, Yousria; Huettmann, Falk; Qu, Fangyuan; Yu, Zishan; Pitcher, C. Roland; Haedrich, Richard L.; Wicksten, Mary K.; Rex, Michael A.; Baguley, Jeffrey G.; Sharma, Jyotsna; Danovaro, Roberto; MacDonald, Ian R.; Nunnally, Clifton C.; Deming, Jody W.; Montagna, Paul; Lévesque, Mélanie; Weslawski, Jan Marcin; Wlodarska-Kowalczuk, Maria; Ingole, Baban S.; Bett, Brian J.; Billett, David S. M.; Yool, Andrew; Bluhm, Bodil A.; Iken, Katrin; Narayanaswamy, Bhavani E.

2010-01-01

A comprehensive seafloor biomass and abundance database has been constructed from 24 oceanographic institutions worldwide within the Census of Marine Life (CoML) field projects. The machine-learning algorithm, Random Forests, was employed to model and predict seafloor standing stocks from surface primary production, water-column integrated and export particulate organic matter (POM), seafloor relief, and bottom water properties. The predictive models explain 63% to 88% of stock variance among the major size groups. Individual and composite maps of predicted global seafloor biomass and abundance are generated for bacteria, meiofauna, macrofauna, and megafauna (invertebrates and fishes). Patterns of benthic standing stocks were positive functions of surface primary production and delivery of the particulate organic carbon (POC) flux to the seafloor. At a regional scale, the census maps illustrate that integrated biomass is highest at the poles, on continental margins associated with coastal upwelling and with broad zones associated with equatorial divergence. Lowest values are consistently encountered on the central abyssal plains of major ocean basins The shift of biomass dominance groups with depth is shown to be affected by the decrease in average body size rather than abundance, presumably due to decrease in quantity and quality of food supply. This biomass census and associated maps are vital components of mechanistic deep-sea food web models and global carbon cycling, and as such provide fundamental information that can be incorporated into evidence-based management. PMID:21209928

Dynamic Resource Allocation and Access Class Barring Scheme for Delay-Sensitive Devices in Machine to Machine (M2M) Communications.

PubMed

Li, Ning; Cao, Chao; Wang, Cong

2017-06-15

Supporting simultaneous access of machine-type devices is a critical challenge in machine-to-machine (M2M) communications. In this paper, we propose an optimal scheme to dynamically adjust the Access Class Barring (ACB) factor and the number of random access channel (RACH) resources for clustered machine-to-machine (M2M) communications, in which Delay-Sensitive (DS) devices coexist with Delay-Tolerant (DT) ones. In M2M communications, since delay-sensitive devices share random access resources with delay-tolerant devices, reducing the resources consumed by delay-sensitive devices means that there will be more resources available to delay-tolerant ones. Our goal is to optimize the random access scheme, which can not only satisfy the requirements of delay-sensitive devices, but also take the communication quality of delay-tolerant ones into consideration. We discuss this problem from the perspective of delay-sensitive services by adjusting the resource allocation and ACB scheme for these devices dynamically. Simulation results show that our proposed scheme realizes good performance in satisfying the delay-sensitive services as well as increasing the utilization rate of the random access resources allocated to them.

Automatic Gleason grading of prostate cancer using SLIM and machine learning

NASA Astrophysics Data System (ADS)

Nguyen, Tan H.; Sridharan, Shamira; Marcias, Virgilia; Balla, Andre K.; Do, Minh N.; Popescu, Gabriel

2016-03-01

In this paper, we present an updated automatic diagnostic procedure for prostate cancer using quantitative phase imaging (QPI). In a recent report [1], we demonstrated the use of Random Forest for image segmentation on prostate cores imaged using QPI. Based on these label maps, we developed an algorithm to discriminate between regions with Gleason grade 3 and 4 prostate cancer in prostatectomy tissue. The Area-Under-Curve (AUC) of 0.79 for the Receiver Operating Curve (ROC) can be obtained for Gleason grade 4 detection in a binary classification between Grade 3 and Grade 4. Our dataset includes 280 benign cases and 141 malignant cases. We show that textural features in phase maps have strong diagnostic values since they can be used in combination with the label map to detect presence or absence of basal cells, which is a strong indicator for prostate carcinoma. A support vector machine (SVM) classifier trained on this new feature vector can classify cancer/non-cancer with an error rate of 0.23 and an AUC value of 0.83.

Predictive analysis of the influence of the chemical composition and pre-processing regimen on structural properties of steel alloys using machine learning techniques

NASA Astrophysics Data System (ADS)

Krishnamurthy, Narayanan; Maddali, Siddharth; Romanov, Vyacheslav; Hawk, Jeffrey

We present some structural properties of multi-component steel alloys as predicted by a random forest machine-learning model. These non-parametric models are trained on high-dimensional data sets defined by features such as chemical composition, pre-processing temperatures and environmental influences, the latter of which are based upon standardized testing procedures for tensile, creep and rupture properties as defined by the American Society of Testing and Materials (ASTM). We quantify the goodness of fit of these models as well as the inferred relative importance of each of these features, all with a conveniently defined metric and scale. The models are tested with synthetic data points, generated subject to the appropriate mathematical constraints for the various features. By this we highlight possible trends in the increase or degradation of the structural properties with perturbations in the features of importance. This work is presented as part of the Data Science Initiative at the National Energy Technology Laboratory, directed specifically towards the computational design of steel alloys.

Is Romantic Desire Predictable? Machine Learning Applied to Initial Romantic Attraction.

PubMed

Joel, Samantha; Eastwick, Paul W; Finkel, Eli J

2017-10-01

Matchmaking companies and theoretical perspectives on close relationships suggest that initial attraction is, to some extent, a product of two people's self-reported traits and preferences. We used machine learning to test how well such measures predict people's overall tendencies to romantically desire other people (actor variance) and to be desired by other people (partner variance), as well as people's desire for specific partners above and beyond actor and partner variance (relationship variance). In two speed-dating studies, romantically unattached individuals completed more than 100 self-report measures about traits and preferences that past researchers have identified as being relevant to mate selection. Each participant met each opposite-sex participant attending a speed-dating event for a 4-min speed date. Random forests models predicted 4% to 18% of actor variance and 7% to 27% of partner variance; crucially, however, they were unable to predict relationship variance using any combination of traits and preferences reported before the dates. These results suggest that compatibility elements of human mating are challenging to predict before two people meet.

Developing a radiomics framework for classifying non-small cell lung carcinoma subtypes

NASA Astrophysics Data System (ADS)

Yu, Dongdong; Zang, Yali; Dong, Di; Zhou, Mu; Gevaert, Olivier; Fang, Mengjie; Shi, Jingyun; Tian, Jie

2017-03-01

Patient-targeted treatment of non-small cell lung carcinoma (NSCLC) has been well documented according to the histologic subtypes over the past decade. In parallel, recent development of quantitative image biomarkers has recently been highlighted as important diagnostic tools to facilitate histological subtype classification. In this study, we present a radiomics analysis that classifies the adenocarcinoma (ADC) and squamous cell carcinoma (SqCC). We extract 52-dimensional, CT-based features (7 statistical features and 45 image texture features) to represent each nodule. We evaluate our approach on a clinical dataset including 324 ADCs and 110 SqCCs patients with CT image scans. Classification of these features is performed with four different machine-learning classifiers including Support Vector Machines with Radial Basis Function kernel (RBF-SVM), Random forest (RF), K-nearest neighbor (KNN), and RUSBoost algorithms. To improve the classifiers' performance, optimal feature subset is selected from the original feature set by using an iterative forward inclusion and backward eliminating algorithm. Extensive experimental results demonstrate that radiomics features achieve encouraging classification results on both complete feature set (AUC=0.89) and optimal feature subset (AUC=0.91).

Gaps in Data and Modeling Tools for Understanding Fire and Fire Effects in Tundra Ecosystems

NASA Astrophysics Data System (ADS)

French, N. H.; Miller, M. E.; Loboda, T. V.; Jenkins, L. K.; Bourgeau-Chavez, L. L.; Suiter, A.; Hawkins, S. M.

2013-12-01

As the ecosystem science community learns more about tundra ecosystems and disturbance in tundra, a review of base data sets and ecological field data for the region shows there are many gaps that need to be filled. In this paper we will review efforts to improve our knowledge of the occurrence and impacts of fire in the North American tundra region completed under a NASA Terrestrial Ecology grant. Our main source of information is remote sensing data from satellite sensors and ecological data from past and recent field data collections by our team, collaborators, and others. Past fire occurrence is not well known for this region compared with other North American biomes. In this presentation we review an effort to use a semi-automated detection algorithm to identify past fire occurrence using the Landsat TM/ETM+ archives, pointing out some of the still-unaddressed issues for a full understanding of fire regime for the region. For this task, fires in Landsat scenes were mapped using the Random Forest classifier (Breiman 2001) to automatically detect potential burn scars. Random Forests is an ensemble classifier that employs machine learning to build a large collection of decision trees that are grown from a random selection of user supplied training data. A pixel's classification is then determined by which class receives the most 'votes' from each tree. We also review the use fire location records and existing modeling methods to quantify emissions from these fires. Based on existing maps of vegetation fuels, we used the approach developed for the Wildland Fire Emissions Information System (WFEIS; French et al. 2011) to estimate emissions across the tundra region. WFEIS employs the Consume model (http://www.fs.fed.us/pnw/fera/research/smoke/consume/index.shtml) to estimate emissions by applying empirically developed relationships between fuels, fire conditions (weather-based fire indexes), and emissions. Here again, we will review the gaps in data and modeling capability for accurate estimation of fire emissions in this region. Initial evaluation of Landsat for tundra fire characterization (Loboda et al. 2013) and successful use of the rich archive of Synthetic Aperture Radar imagery for many fire-disturbed sites in the region will be additional topics covered in this poster presentation. References: Breiman, L. 2001. Random forests. Machine Learning, 45:5-32. French, N.H.F., W.J. de Groot, L.K. Jenkins, B.. Rogers, et al. 2011. Model comparisons for estimating carbon emissions from North American wildland fire. J. Geophys. Res. 116:G00K05, doi:10.1029/2010JG001469. Loboda, T L, N H F French, C. Hight-Harf, L. Jenkins, M.E. Miller. 2013. Mapping fire extent and burn severity in Alaskan tussock tundra: An analysis of the spectral response of tundra vegetation to wildland fire. Remote Sens. Enviro. 134:194-209.

Predicting surface fuel models and fuel metrics using lidar and CIR imagery in a dense mixed conifer forest

Treesearch

Marek K. Jakubowksi; Qinghua Guo; Brandon Collins; Scott Stephens; Maggi Kelly

2013-01-01

We compared the ability of several classification and regression algorithms to predict forest stand structure metrics and standard surface fuel models. Our study area spans a dense, topographically complex Sierra Nevada mixed-conifer forest. We used clustering, regression trees, and support vector machine algorithms to analyze high density (average 9 pulses/m

Microgravity simulations with human lymphocytes in the free fall machine and in the random positioning machine

NASA Technical Reports Server (NTRS)

Schwarzenberg, M.; Pippia, P.; Meloni, M. A.; Cossu, G.; Cogoli-Greuter, M.; Cogoli, A.

1998-01-01

The purpose of this paper is to present the results obtained in our laboratory with both instruments, the FFM [free fall machine] and the RPM [random positioning machine], to compare them with the data from earlier experiments with human lymphocytes conducted in the FRC [fast rotating clinostat] and in space. Furthermore, the suitability of the FFM and RPM for research in gravitational cell biology is discussed.

Applying genetic algorithms to set the optimal combination of forest fire related variables and model forest fire susceptibility based on data mining models. The case of Dayu County, China.

PubMed

Hong, Haoyuan; Tsangaratos, Paraskevas; Ilia, Ioanna; Liu, Junzhi; Zhu, A-Xing; Xu, Chong

2018-07-15

The main objective of the present study was to utilize Genetic Algorithms (GA) in order to obtain the optimal combination of forest fire related variables and apply data mining methods for constructing a forest fire susceptibility map. In the proposed approach, a Random Forest (RF) and a Support Vector Machine (SVM) was used to produce a forest fire susceptibility map for the Dayu County which is located in southwest of Jiangxi Province, China. For this purpose, historic forest fires and thirteen forest fire related variables were analyzed, namely: elevation, slope angle, aspect, curvature, land use, soil cover, heat load index, normalized difference vegetation index, mean annual temperature, mean annual wind speed, mean annual rainfall, distance to river network and distance to road network. The Natural Break and the Certainty Factor method were used to classify and weight the thirteen variables, while a multicollinearity analysis was performed to determine the correlation among the variables and decide about their usability. The optimal set of variables, determined by the GA limited the number of variables into eight excluding from the analysis, aspect, land use, heat load index, distance to river network and mean annual rainfall. The performance of the forest fire models was evaluated by using the area under the Receiver Operating Characteristic curve (ROC-AUC) based on the validation dataset. Overall, the RF models gave higher AUC values. Also the results showed that the proposed optimized models outperform the original models. Specifically, the optimized RF model gave the best results (0.8495), followed by the original RF (0.8169), while the optimized SVM gave lower values (0.7456) than the RF, however higher than the original SVM (0.7148) model. The study highlights the significance of feature selection techniques in forest fire susceptibility, whereas data mining methods could be considered as a valid approach for forest fire susceptibility modeling. Copyright © 2018 Elsevier B.V. All rights reserved.

Genome-wide association data classification and SNPs selection using two-stage quality-based Random Forests.

PubMed

Nguyen, Thanh-Tung; Huang, Joshua; Wu, Qingyao; Nguyen, Thuy; Li, Mark

2015-01-01

Single-nucleotide polymorphisms (SNPs) selection and identification are the most important tasks in Genome-wide association data analysis. The problem is difficult because genome-wide association data is very high dimensional and a large portion of SNPs in the data is irrelevant to the disease. Advanced machine learning methods have been successfully used in Genome-wide association studies (GWAS) for identification of genetic variants that have relatively big effects in some common, complex diseases. Among them, the most successful one is Random Forests (RF). Despite of performing well in terms of prediction accuracy in some data sets with moderate size, RF still suffers from working in GWAS for selecting informative SNPs and building accurate prediction models. In this paper, we propose to use a new two-stage quality-based sampling method in random forests, named ts-RF, for SNP subspace selection for GWAS. The method first applies p-value assessment to find a cut-off point that separates informative and irrelevant SNPs in two groups. The informative SNPs group is further divided into two sub-groups: highly informative and weak informative SNPs. When sampling the SNP subspace for building trees for the forest, only those SNPs from the two sub-groups are taken into account. The feature subspaces always contain highly informative SNPs when used to split a node at a tree. This approach enables one to generate more accurate trees with a lower prediction error, meanwhile possibly avoiding overfitting. It allows one to detect interactions of multiple SNPs with the diseases, and to reduce the dimensionality and the amount of Genome-wide association data needed for learning the RF model. Extensive experiments on two genome-wide SNP data sets (Parkinson case-control data comprised of 408,803 SNPs and Alzheimer case-control data comprised of 380,157 SNPs) and 10 gene data sets have demonstrated that the proposed model significantly reduced prediction errors and outperformed most existing the-state-of-the-art random forests. The top 25 SNPs in Parkinson data set were identified by the proposed model including four interesting genes associated with neurological disorders. The presented approach has shown to be effective in selecting informative sub-groups of SNPs potentially associated with diseases that traditional statistical approaches might fail. The new RF works well for the data where the number of case-control objects is much smaller than the number of SNPs, which is a typical problem in gene data and GWAS. Experiment results demonstrated the effectiveness of the proposed RF model that outperformed the state-of-the-art RFs, including Breiman's RF, GRRF and wsRF methods.

An automated ranking platform for machine learning regression models for meat spoilage prediction using multi-spectral imaging and metabolic profiling.

PubMed

Estelles-Lopez, Lucia; Ropodi, Athina; Pavlidis, Dimitris; Fotopoulou, Jenny; Gkousari, Christina; Peyrodie, Audrey; Panagou, Efstathios; Nychas, George-John; Mohareb, Fady

2017-09-01

Over the past decade, analytical approaches based on vibrational spectroscopy, hyperspectral/multispectral imagining and biomimetic sensors started gaining popularity as rapid and efficient methods for assessing food quality, safety and authentication; as a sensible alternative to the expensive and time-consuming conventional microbiological techniques. Due to the multi-dimensional nature of the data generated from such analyses, the output needs to be coupled with a suitable statistical approach or machine-learning algorithms before the results can be interpreted. Choosing the optimum pattern recognition or machine learning approach for a given analytical platform is often challenging and involves a comparative analysis between various algorithms in order to achieve the best possible prediction accuracy. In this work, "MeatReg", a web-based application is presented, able to automate the procedure of identifying the best machine learning method for comparing data from several analytical techniques, to predict the counts of microorganisms responsible of meat spoilage regardless of the packaging system applied. In particularly up to 7 regression methods were applied and these are ordinary least squares regression, stepwise linear regression, partial least square regression, principal component regression, support vector regression, random forest and k-nearest neighbours. MeatReg" was tested with minced beef samples stored under aerobic and modified atmosphere packaging and analysed with electronic nose, HPLC, FT-IR, GC-MS and Multispectral imaging instrument. Population of total viable count, lactic acid bacteria, pseudomonads, Enterobacteriaceae and B. thermosphacta, were predicted. As a result, recommendations of which analytical platforms are suitable to predict each type of bacteria and which machine learning methods to use in each case were obtained. The developed system is accessible via the link: www.sorfml.com. Copyright © 2017 Elsevier Ltd. All rights reserved.

Comparison of Random Forest and Support Vector Machine classifiers using UAV remote sensing imagery

NASA Astrophysics Data System (ADS)

Piragnolo, Marco; Masiero, Andrea; Pirotti, Francesco

2017-04-01

Since recent years surveying with unmanned aerial vehicles (UAV) is getting a great amount of attention due to decreasing costs, higher precision and flexibility of usage. UAVs have been applied for geomorphological investigations, forestry, precision agriculture, cultural heritage assessment and for archaeological purposes. It can be used for land use and land cover classification (LULC). In literature, there are two main types of approaches for classification of remote sensing imagery: pixel-based and object-based. On one hand, pixel-based approach mostly uses training areas to define classes and respective spectral signatures. On the other hand, object-based classification considers pixels, scale, spatial information and texture information for creating homogeneous objects. Machine learning methods have been applied successfully for classification, and their use is increasing due to the availability of faster computing capabilities. The methods learn and train the model from previous computation. Two machine learning methods which have given good results in previous investigations are Random Forest (RF) and Support Vector Machine (SVM). The goal of this work is to compare RF and SVM methods for classifying LULC using images collected with a fixed wing UAV. The processing chain regarding classification uses packages in R, an open source scripting language for data analysis, which provides all necessary algorithms. The imagery was acquired and processed in November 2015 with cameras providing information over the red, blue, green and near infrared wavelength reflectivity over a testing area in the campus of Agripolis, in Italy. Images were elaborated and ortho-rectified through Agisoft Photoscan. The ortho-rectified image is the full data set, and the test set is derived from partial sub-setting of the full data set. Different tests have been carried out, using a percentage from 2 % to 20 % of the total. Ten training sets and ten validation sets are obtained from each test set. The control dataset consist of an independent visual classification done by an expert over the whole area. The classes are (i) broadleaf, (ii) building, (iii) grass, (iv) headland access path, (v) road, (vi) sowed land, (vii) vegetable. The RF and SVM are applied to the test set. The performances of the methods are evaluated using the three following accuracy metrics: Kappa index, Classification accuracy and Classification Error. All three are calculated in three different ways: with K-fold cross validation, using the validation test set and using the full test set. The analysis indicates that SVM gets better results in terms of good scores using K-fold cross or validation test set. Using the full test set, RF achieves a better result in comparison to SVM. It also seems that SVM performs better with smaller training sets, whereas RF performs better as training sets get larger.

The assisted prediction modelling frame with hybridisation and ensemble for business risk forecasting and an implementation

NASA Astrophysics Data System (ADS)

Li, Hui; Hong, Lu-Yao; Zhou, Qing; Yu, Hai-Jie

2015-08-01

The business failure of numerous companies results in financial crises. The high social costs associated with such crises have made people to search for effective tools for business risk prediction, among which, support vector machine is very effective. Several modelling means, including single-technique modelling, hybrid modelling, and ensemble modelling, have been suggested in forecasting business risk with support vector machine. However, existing literature seldom focuses on the general modelling frame for business risk prediction, and seldom investigates performance differences among different modelling means. We reviewed researches on forecasting business risk with support vector machine, proposed the general assisted prediction modelling frame with hybridisation and ensemble (APMF-WHAE), and finally, investigated the use of principal components analysis, support vector machine, random sampling, and group decision, under the general frame in forecasting business risk. Under the APMF-WHAE frame with support vector machine as the base predictive model, four specific predictive models were produced, namely, pure support vector machine, a hybrid support vector machine involved with principal components analysis, a support vector machine ensemble involved with random sampling and group decision, and an ensemble of hybrid support vector machine using group decision to integrate various hybrid support vector machines on variables produced from principle components analysis and samples from random sampling. The experimental results indicate that hybrid support vector machine and ensemble of hybrid support vector machines were able to produce dominating performance than pure support vector machine and support vector machine ensemble.

Non-random species loss in a forest herbaceous layer following nitrogen addition

Treesearch

Christopher A. ​Walter; Mary Beth Adams; Frank S. Gilliam; William T. Peterjohn

2017-01-01

Nitrogen (N) additions have decreased species richness (S) in hardwood forest herbaceous layers, yet the functional mechanisms for these decreases have not been explicitly evaluated.We tested two hypothesized mechanisms, random species loss (RSL) and non-random species loss (NRSL), in the hardwood forest herbaceous layer of a long-term, plot-scale...

Development of a Computer Vision Technology for the Forest Products Manufacturing Industry

Treesearch

D. Earl Kline; Richard Conners; Philip A. Araman

1992-01-01

The goal of this research is to create an automated processing/grading system for hardwood lumber that will be of use to the forest products industry. The objective of creating a full scale machine vision prototype for inspecting hardwood lumber will become a reality in calendar year 1992. Space for the full scale prototype has been created at the Brooks Forest...

Discrimination of crop types with TerraSAR-X-derived information

NASA Astrophysics Data System (ADS)

Sonobe, Rei; Tani, Hiroshi; Wang, Xiufeng; Kobayashi, Nobuyuki; Shimamura, Hideki

Although classification maps are required for management and for the estimation of agricultural disaster compensation, those techniques have yet to be established. This paper describes the comparison of three different classification algorithms for mapping crops in Hokkaido, Japan, using TerraSAR-X (including TanDEM-X) dual-polarimetric data. In the study area, beans, beets, grasslands, maize, potatoes and winter wheat were cultivated. In this study, classification using TerraSAR-X-derived information was performed. Coherence values, polarimetric parameters and gamma nought values were also obtained and evaluated regarding their usefulness in crop classification. Accurate classification may be possible with currently existing supervised learning models. A comparison between the classification and regression tree (CART), support vector machine (SVM) and random forests (RF) algorithms was performed. Even though J-M distances were lower than 1.0 on all TerraSAR-X acquisition days, good results were achieved (e.g., separability between winter wheat and grass) due to the characteristics of the machine learning algorithm. It was found that SVM performed best, achieving an overall accuracy of 95.0% based on the polarimetric parameters and gamma nought values for HH and VV polarizations. The misclassified fields were less than 100 a in area and 79.5-96.3% were less than 200 a with the exception of grassland. When some feature such as a road or windbreak forest is present in the TerraSAR-X data, the ratio of its extent to that of the field is relatively higher for the smaller fields, which leads to misclassifications.

Predicting temperate forest stand types using only structural profiles from discrete return airborne lidar

NASA Astrophysics Data System (ADS)

Fedrigo, Melissa; Newnham, Glenn J.; Coops, Nicholas C.; Culvenor, Darius S.; Bolton, Douglas K.; Nitschke, Craig R.

2018-02-01

Light detection and ranging (lidar) data have been increasingly used for forest classification due to its ability to penetrate the forest canopy and provide detail about the structure of the lower strata. In this study we demonstrate forest classification approaches using airborne lidar data as inputs to random forest and linear unmixing classification algorithms. Our results demonstrated that both random forest and linear unmixing models identified a distribution of rainforest and eucalypt stands that was comparable to existing ecological vegetation class (EVC) maps based primarily on manual interpretation of high resolution aerial imagery. Rainforest stands were also identified in the region that have not previously been identified in the EVC maps. The transition between stand types was better characterised by the random forest modelling approach. In contrast, the linear unmixing model placed greater emphasis on field plots selected as endmembers which may not have captured the variability in stand structure within a single stand type. The random forest model had the highest overall accuracy (84%) and Cohen's kappa coefficient (0.62). However, the classification accuracy was only marginally better than linear unmixing. The random forest model was applied to a region in the Central Highlands of south-eastern Australia to produce maps of stand type probability, including areas of transition (the 'ecotone') between rainforest and eucalypt forest. The resulting map provided a detailed delineation of forest classes, which specifically recognised the coalescing of stand types at the landscape scale. This represents a key step towards mapping the structural and spatial complexity of these ecosystems, which is important for both their management and conservation.

Promoting the Purchase of Low-Calorie Foods from School Vending Machines: A Cluster-Randomized Controlled Study

ERIC Educational Resources Information Center

Kocken, Paul L.; Eeuwijk, Jennifer; van Kesteren, Nicole M.C.; Dusseldorp, Elise; Buijs, Goof; Bassa-Dafesh, Zeina; Snel, Jeltje

2012-01-01

Background: Vending machines account for food sales and revenue in schools. We examined 3 strategies for promoting the sale of lower-calorie food products from vending machines in high schools in the Netherlands. Methods: A school-based randomized controlled trial was conducted in 13 experimental schools and 15 control schools. Three strategies…

Personalized Physical Activity Coaching: A Machine Learning Approach

PubMed Central

Dijkhuis, Talko B.; van Ittersum, Miriam W.; Velthuijsen, Hugo

2018-01-01

Living a sedentary lifestyle is one of the major causes of numerous health problems. To encourage employees to lead a less sedentary life, the Hanze University started a health promotion program. One of the interventions in the program was the use of an activity tracker to record participants' daily step count. The daily step count served as input for a fortnightly coaching session. In this paper, we investigate the possibility of automating part of the coaching procedure on physical activity by providing personalized feedback throughout the day on a participant’s progress in achieving a personal step goal. The gathered step count data was used to train eight different machine learning algorithms to make hourly estimations of the probability of achieving a personalized, daily steps threshold. In 80% of the individual cases, the Random Forest algorithm was the best performing algorithm (mean accuracy = 0.93, range = 0.88–0.99, and mean F1-score = 0.90, range = 0.87–0.94). To demonstrate the practical usefulness of these models, we developed a proof-of-concept Web application that provides personalized feedback about whether a participant is expected to reach his or her daily threshold. We argue that the use of machine learning could become an invaluable asset in the process of automated personalized coaching. The individualized algorithms allow for predicting physical activity during the day and provides the possibility to intervene in time. PMID:29463052

In silico prediction of ROCK II inhibitors by different classification approaches.

PubMed

Cai, Chuipu; Wu, Qihui; Luo, Yunxia; Ma, Huili; Shen, Jiangang; Zhang, Yongbin; Yang, Lei; Chen, Yunbo; Wen, Zehuai; Wang, Qi

2017-11-01

ROCK II is an important pharmacological target linked to central nervous system disorders such as Alzheimer's disease. The purpose of this research is to generate ROCK II inhibitor prediction models by machine learning approaches. Firstly, four sets of descriptors were calculated with MOE 2010 and PaDEL-Descriptor, and optimized by F-score and linear forward selection methods. In addition, four classification algorithms were used to initially build 16 classifiers with k-nearest neighbors [Formula: see text], naïve Bayes, Random forest, and support vector machine. Furthermore, three sets of structural fingerprint descriptors were introduced to enhance the predictive capacity of classifiers, which were assessed with fivefold cross-validation, test set validation and external test set validation. The best two models, MFK + MACCS and MLR + SubFP, have both MCC values of 0.925 for external test set. After that, a privileged substructure analysis was performed to reveal common chemical features of ROCK II inhibitors. Finally, binding modes were analyzed to identify relationships between molecular descriptors and activity, while main interactions were revealed by comparing the docking interaction of the most potent and the weakest ROCK II inhibitors. To the best of our knowledge, this is the first report on ROCK II inhibitors utilizing machine learning approaches that provides a new method for discovering novel ROCK II inhibitors.

Type 2 Diabetes Screening Test by Means of a Pulse Oximeter.

PubMed

Moreno, Enrique Monte; Lujan, Maria Jose Anyo; Rusinol, Montse Torrres; Fernandez, Paqui Juarez; Manrique, Pilar Nunez; Trivino, Cristina Aragon; Miquel, Magda Pedrosa; Rodriguez, Marife Alvarez; Burguillos, M Jose Gonzalez

2017-02-01

In this paper, we propose a method for screening for the presence of type 2 diabetes by means of the signal obtained from a pulse oximeter. The screening system consists of two parts: the first analyzes the signal obtained from the pulse oximeter, and the second consists of a machine-learning module. The system consists of a front end that extracts a set of features form the pulse oximeter signal. These features are based on physiological considerations. The set of features were the input of a machine-learning algorithm that determined the class of the input sample, i.e., whether the subject had diabetes or not. The machine-learning algorithms were random forests, gradient boosting, and linear discriminant analysis as benchmark. The system was tested on a database of [Formula: see text] subjects (two samples per subject) collected from five community health centers. The mean receiver operating characteristic area found was [Formula: see text]% (median value [Formula: see text]% and range [Formula: see text]%), with a specificity =  [Formula: see text]% for a threshold that gave a sensitivity = [Formula: see text]%. We present a screening method for detecting diabetes that has a performance comparable to the glycated haemoglobin (haemoglobin A1c HbA1c) test, does not require blood extraction, and yields results in less than 5 min.

A Machine Learning Application Based in Random Forest for Integrating Mass Spectrometry-Based Metabolomic Data: A Simple Screening Method for Patients With Zika Virus

PubMed Central

Melo, Carlos Fernando Odir Rodrigues; Navarro, Luiz Claudio; de Oliveira, Diogo Noin; Guerreiro, Tatiane Melina; Lima, Estela de Oliveira; Delafiori, Jeany; Dabaja, Mohamed Ziad; Ribeiro, Marta da Silva; de Menezes, Maico; Rodrigues, Rafael Gustavo Martins; Morishita, Karen Noda; Esteves, Cibele Zanardi; de Amorim, Aline Lopes Lucas; Aoyagui, Caroline Tiemi; Parise, Pierina Lorencini; Milanez, Guilherme Paier; do Nascimento, Gabriela Mansano; Ribas Freitas, André Ricardo; Angerami, Rodrigo; Costa, Fábio Trindade Maranhão; Arns, Clarice Weis; Resende, Mariangela Ribeiro; Amaral, Eliana; Junior, Renato Passini; Ribeiro-do-Valle, Carolina C.; Milanez, Helaine; Moretti, Maria Luiza; Proenca-Modena, Jose Luiz; Avila, Sandra; Rocha, Anderson; Catharino, Rodrigo Ramos

2018-01-01

Recent Zika outbreaks in South America, accompanied by unexpectedly severe clinical complications have brought much interest in fast and reliable screening methods for ZIKV (Zika virus) identification. Reverse-transcriptase polymerase chain reaction (RT-PCR) is currently the method of choice to detect ZIKV in biological samples. This approach, nonetheless, demands a considerable amount of time and resources such as kits and reagents that, in endemic areas, may result in a substantial financial burden over affected individuals and health services veering away from RT-PCR analysis. This study presents a powerful combination of high-resolution mass spectrometry and a machine-learning prediction model for data analysis to assess the existence of ZIKV infection across a series of patients that bear similar symptomatic conditions, but not necessarily are infected with the disease. By using mass spectrometric data that are inputted with the developed decision-making algorithm, we were able to provide a set of features that work as a “fingerprint” for this specific pathophysiological condition, even after the acute phase of infection. Since both mass spectrometry and machine learning approaches are well-established and have largely utilized tools within their respective fields, this combination of methods emerges as a distinct alternative for clinical applications, providing a diagnostic screening—faster and more accurate—with improved cost-effectiveness when compared to existing technologies. PMID:29696139

A Machine Learning Application Based in Random Forest for Integrating Mass Spectrometry-Based Metabolomic Data: A Simple Screening Method for Patients With Zika Virus.

PubMed

Melo, Carlos Fernando Odir Rodrigues; Navarro, Luiz Claudio; de Oliveira, Diogo Noin; Guerreiro, Tatiane Melina; Lima, Estela de Oliveira; Delafiori, Jeany; Dabaja, Mohamed Ziad; Ribeiro, Marta da Silva; de Menezes, Maico; Rodrigues, Rafael Gustavo Martins; Morishita, Karen Noda; Esteves, Cibele Zanardi; de Amorim, Aline Lopes Lucas; Aoyagui, Caroline Tiemi; Parise, Pierina Lorencini; Milanez, Guilherme Paier; do Nascimento, Gabriela Mansano; Ribas Freitas, André Ricardo; Angerami, Rodrigo; Costa, Fábio Trindade Maranhão; Arns, Clarice Weis; Resende, Mariangela Ribeiro; Amaral, Eliana; Junior, Renato Passini; Ribeiro-do-Valle, Carolina C; Milanez, Helaine; Moretti, Maria Luiza; Proenca-Modena, Jose Luiz; Avila, Sandra; Rocha, Anderson; Catharino, Rodrigo Ramos

2018-01-01

Recent Zika outbreaks in South America, accompanied by unexpectedly severe clinical complications have brought much interest in fast and reliable screening methods for ZIKV (Zika virus) identification. Reverse-transcriptase polymerase chain reaction (RT-PCR) is currently the method of choice to detect ZIKV in biological samples. This approach, nonetheless, demands a considerable amount of time and resources such as kits and reagents that, in endemic areas, may result in a substantial financial burden over affected individuals and health services veering away from RT-PCR analysis. This study presents a powerful combination of high-resolution mass spectrometry and a machine-learning prediction model for data analysis to assess the existence of ZIKV infection across a series of patients that bear similar symptomatic conditions, but not necessarily are infected with the disease. By using mass spectrometric data that are inputted with the developed decision-making algorithm, we were able to provide a set of features that work as a "fingerprint" for this specific pathophysiological condition, even after the acute phase of infection. Since both mass spectrometry and machine learning approaches are well-established and have largely utilized tools within their respective fields, this combination of methods emerges as a distinct alternative for clinical applications, providing a diagnostic screening-faster and more accurate-with improved cost-effectiveness when compared to existing technologies.

An automated, high-throughput plant phenotyping system using machine learning-based plant segmentation and image analysis.

PubMed

Lee, Unseok; Chang, Sungyul; Putra, Gian Anantrio; Kim, Hyoungseok; Kim, Dong Hwan

2018-01-01

A high-throughput plant phenotyping system automatically observes and grows many plant samples. Many plant sample images are acquired by the system to determine the characteristics of the plants (populations). Stable image acquisition and processing is very important to accurately determine the characteristics. However, hardware for acquiring plant images rapidly and stably, while minimizing plant stress, is lacking. Moreover, most software cannot adequately handle large-scale plant imaging. To address these problems, we developed a new, automated, high-throughput plant phenotyping system using simple and robust hardware, and an automated plant-imaging-analysis pipeline consisting of machine-learning-based plant segmentation. Our hardware acquires images reliably and quickly and minimizes plant stress. Furthermore, the images are processed automatically. In particular, large-scale plant-image datasets can be segmented precisely using a classifier developed using a superpixel-based machine-learning algorithm (Random Forest), and variations in plant parameters (such as area) over time can be assessed using the segmented images. We performed comparative evaluations to identify an appropriate learning algorithm for our proposed system, and tested three robust learning algorithms. We developed not only an automatic analysis pipeline but also a convenient means of plant-growth analysis that provides a learning data interface and visualization of plant growth trends. Thus, our system allows end-users such as plant biologists to analyze plant growth via large-scale plant image data easily.

Genome-wide prediction and analysis of human tissue-selective genes using microarray expression data

PubMed Central

2013-01-01

Background Understanding how genes are expressed specifically in particular tissues is a fundamental question in developmental biology. Many tissue-specific genes are involved in the pathogenesis of complex human diseases. However, experimental identification of tissue-specific genes is time consuming and difficult. The accurate predictions of tissue-specific gene targets could provide useful information for biomarker development and drug target identification. Results In this study, we have developed a machine learning approach for predicting the human tissue-specific genes using microarray expression data. The lists of known tissue-specific genes for different tissues were collected from UniProt database, and the expression data retrieved from the previously compiled dataset according to the lists were used for input vector encoding. Random Forests (RFs) and Support Vector Machines (SVMs) were used to construct accurate classifiers. The RF classifiers were found to outperform SVM models for tissue-specific gene prediction. The results suggest that the candidate genes for brain or liver specific expression can provide valuable information for further experimental studies. Our approach was also applied for identifying tissue-selective gene targets for different types of tissues. Conclusions A machine learning approach has been developed for accurately identifying the candidate genes for tissue specific/selective expression. The approach provides an efficient way to select some interesting genes for developing new biomedical markers and improve our knowledge of tissue-specific expression. PMID:23369200

Scalable Nearest Neighbor Algorithms for High Dimensional Data.

PubMed

Muja, Marius; Lowe, David G

2014-11-01

For many computer vision and machine learning problems, large training sets are key for good performance. However, the most computationally expensive part of many computer vision and machine learning algorithms consists of finding nearest neighbor matches to high dimensional vectors that represent the training data. We propose new algorithms for approximate nearest neighbor matching and evaluate and compare them with previous algorithms. For matching high dimensional features, we find two algorithms to be the most efficient: the randomized k-d forest and a new algorithm proposed in this paper, the priority search k-means tree. We also propose a new algorithm for matching binary features by searching multiple hierarchical clustering trees and show it outperforms methods typically used in the literature. We show that the optimal nearest neighbor algorithm and its parameters depend on the data set characteristics and describe an automated configuration procedure for finding the best algorithm to search a particular data set. In order to scale to very large data sets that would otherwise not fit in the memory of a single machine, we propose a distributed nearest neighbor matching framework that can be used with any of the algorithms described in the paper. All this research has been released as an open source library called fast library for approximate nearest neighbors (FLANN), which has been incorporated into OpenCV and is now one of the most popular libraries for nearest neighbor matching.

A comparison of machine learning techniques for survival prediction in breast cancer

PubMed Central

2011-01-01

Background The ability to accurately classify cancer patients into risk classes, i.e. to predict the outcome of the pathology on an individual basis, is a key ingredient in making therapeutic decisions. In recent years gene expression data have been successfully used to complement the clinical and histological criteria traditionally used in such prediction. Many "gene expression signatures" have been developed, i.e. sets of genes whose expression values in a tumor can be used to predict the outcome of the pathology. Here we investigate the use of several machine learning techniques to classify breast cancer patients using one of such signatures, the well established 70-gene signature. Results We show that Genetic Programming performs significantly better than Support Vector Machines, Multilayered Perceptrons and Random Forests in classifying patients from the NKI breast cancer dataset, and comparably to the scoring-based method originally proposed by the authors of the 70-gene signature. Furthermore, Genetic Programming is able to perform an automatic feature selection. Conclusions Since the performance of Genetic Programming is likely to be improvable compared to the out-of-the-box approach used here, and given the biological insight potentially provided by the Genetic Programming solutions, we conclude that Genetic Programming methods are worth further investigation as a tool for cancer patient classification based on gene expression data. PMID:21569330

A method to combine target volume data from 3D and 4D planned thoracic radiotherapy patient cohorts for machine learning applications.

PubMed

Johnson, Corinne; Price, Gareth; Khalifa, Jonathan; Faivre-Finn, Corinne; Dekker, Andre; Moore, Christopher; van Herk, Marcel

2018-02-01

The gross tumour volume (GTV) is predictive of clinical outcome and consequently features in many machine-learned models. 4D-planning, however, has prompted substitution of the GTV with the internal gross target volume (iGTV). We present and validate a method to synthesise GTV data from the iGTV, allowing the combination of 3D and 4D planned patient cohorts for modelling. Expert delineations in 40 non-small cell lung cancer patients were used to develop linear fit and erosion methods to synthesise the GTV volume and shape. Quality was assessed using Dice Similarity Coefficients (DSC) and closest point measurements; by calculating dosimetric features; and by assessing the quality of random forest models built on patient populations with and without synthetic GTVs. Volume estimates were within the magnitudes of inter-observer delineation variability. Shape comparisons produced mean DSCs of 0.8817 and 0.8584 for upper and lower lobe cases, respectively. A model trained on combined true and synthetic data performed significantly better than models trained on GTV alone, or combined GTV and iGTV data. Accurate synthesis of GTV size from the iGTV permits the combination of lung cancer patient cohorts, facilitating machine learning applications in thoracic radiotherapy. Copyright © 2017 Elsevier B.V. All rights reserved.

Disruption Warning Database Development and Exploratory Machine Learning Studies on Alcator C-Mod

NASA Astrophysics Data System (ADS)

Montes, Kevin; Rea, Cristina; Granetz, Robert

2017-10-01

A database of about 1800 shots from the 2015 campaign on the Alcator C-Mod tokamak is assembled, including disruptive and non-disruptive discharges. The database consists of 40 relevant plasma parameters with data taken from 160k time slices. In order to investigate the possibility of developing a robust disruption prediction algorithm that is tokamak-independent, we focused machine learning studies on a subset of dimensionless parameters such as βp, n /nG , etc. The Random Forests machine learning algorithm provides insight on the available data set by ranking the relative importance of the input features. Its application on the C-Mod database, however, reveals that virtually no one parameter has more importance than any other, and that its classification algorithm has a low rate of successfully predicted samples, as well as poor false positive and false negative rates. Comparing the analysis of this algorithm on the C-Mod database with its application to a similar database on DIII-D, we conclude that disruption prediction may not be feasible on C-Mod. This conclusion is supported by empirical observations that most C-Mod disruptions are caused by radiative collapse due to molybdenum from the first wall, which happens on just a 1-2ms timescale. Supported by the US Dept. of Energy under DE-FC02-99ER54512 and DE-FC02-04ER54698.

The experimental design of the Missouri Ozark Forest Ecosystem Project

Treesearch

Steven L. Sheriff; Shuoqiong He

1997-01-01

The Missouri Ozark Forest Ecosystem Project (MOFEP) is an experiment that examines the effects of three forest management practices on the forest community. MOFEP is designed as a randomized complete block design using nine sites divided into three blocks. Treatments of uneven-aged, even-aged, and no-harvest management were randomly assigned to sites within each block...

Application of machine learning methods to describe the effects of conjugated equine estrogens therapy on region-specific brain volumes.

PubMed

Casanova, Ramon; Espeland, Mark A; Goveas, Joseph S; Davatzikos, Christos; Gaussoin, Sarah A; Maldjian, Joseph A; Brunner, Robert L; Kuller, Lewis H; Johnson, Karen C; Mysiw, W Jerry; Wagner, Benjamin; Resnick, Susan M

2011-05-01

Use of conjugated equine estrogens (CEE) has been linked to smaller regional brain volumes in women aged ≥65 years; however, it is unknown whether this results in a broad-based characteristic pattern of effects. Structural magnetic resonance imaging was used to assess regional volumes of normal tissue and ischemic lesions among 513 women who had been enrolled in a randomized clinical trial of CEE therapy for an average of 6.6 years, beginning at ages 65-80 years. A multivariate pattern analysis, based on a machine learning technique that combined Random Forest and logistic regression with L(1) penalty, was applied to identify patterns among regional volumes associated with therapy and whether patterns discriminate between treatment groups. The multivariate pattern analysis detected smaller regional volumes of normal tissue within the limbic and temporal lobes among women that had been assigned to CEE therapy. Mean decrements ranged as high as 7% in the left entorhinal cortex and 5% in the left perirhinal cortex, which exceeded the effect sizes reported previously in frontal lobe and hippocampus. Overall accuracy of classification based on these patterns, however, was projected to be only 54.5%. Prescription of CEE therapy for an average of 6.6 years is associated with lower regional brain volumes, but it does not induce a characteristic spatial pattern of changes in brain volumes of sufficient magnitude to discriminate users and nonusers. Copyright © 2011 Elsevier Inc. All rights reserved.

Application of machine learning methods to describe the effects of conjugated equine estrogens therapy on region-specific brain volumes

PubMed Central

Casanova, Ramon; Espeland, Mark A.; Goveas, Joseph S.; Davatzikos, Christos; Gaussoin, Sarah A.; Maldjian, Joseph A.; Brunner, Robert L.; Kuller, Lewis H.; Johnson, Karen C.; Mysiw, W. Jerry; Wagner, Benjamin; Resnick, Susan M.

2011-01-01

Use of conjugated equine estrogens (CEE) has been linked to smaller regional brain volumes in women aged ≥65 years, however it is unknown whether this results in a broad-based characteristic pattern of effects. Structural MRI was used to assess regional volumes of normal tissue and ischemic lesions among 513 women who had been enrolled in a randomized clinical trial of CEE therapy for an average of 6.6 years, beginning at ages 65-80 years. A multivariate pattern analysis, based on a machine learning technique that combined Random Forest and logistic regression with L1 penalty, was applied to identify patterns among regional volumes associated with therapy and whether patterns discriminate between treatment groups. The multivariate pattern analysis detected smaller regional volumes of normal tissue within the limbic and temporal lobes among women that had been assigned to CEE therapy. Mean decrements ranged as high as 7% in the left entorhinal cortex and 5% in the left perirhinal cortex, which exceeded the effect sizes reported previously in frontal lobe and hippocampus. Overall accuracy of classification based on these patterns, however, was projected to be only 54.5%. Prescription of CEE therapy for an average of 6.6 years is associated with lower regional brain volumes, but it does not induce a characteristic spatial pattern of changes in brain volumes of sufficient magnitude to discriminate users and non-users. PMID:21292420

Assessing the Influence of Precipitation Variability on the Vegetation Dynamics of the Mediterranean Rangelands using NDVI and Machine Learning

NASA Astrophysics Data System (ADS)

Daliakopoulos, Ioannis; Tsanis, Ioannis

2017-04-01

Mitigating the vulnerability of Mediterranean rangelands against degradation is limited by our ability to understand and accurately characterize those impacts in space and time. The Normalized Difference Vegetation Index (NDVI) is a radiometric measure of the photosynthetically active radiation absorbed by green vegetation canopy chlorophyll and is therefore a good surrogate measure of vegetation dynamics. On the other hand, meteorological indices such as the drought assessing Standardised Precipitation Index (SPI) are can be easily estimated from historical and projected datasets at the global scale. This work investigates the potential of driving Random Forest (RF) models with meteorological indices to approximate NDVI-based vegetation dynamics. A sufficiently large number of RF models are trained using random subsets of the dataset as predictors, in a bootstrapping approach to account for the uncertainty introduced by the subset selection. The updated E-OBS-v13.1 dataset of the ENSEMBLES EU FP6 program provides observed monthly meteorological input to estimate SPI over the Mediterranean rangelands. RF models are trained to depict vegetation dynamics using the latest version (3g.v1) of the third generation GIMMS NDVI generated from NOAA's Advanced Very High Resolution Radiometer (AVHRR) sensors. Analysis is conducted for the period 1981-2015 at a gridded spatial resolution of 25 km. Preliminary results demonstrate the potential of machine learning algorithms to effectively mimic the underlying physical relationship of drought and Earth Observation vegetation indices to provide estimates based on precipitation variability.

Sci-Fri AM: Quality, Safety, and Professional Issues 04: Predicting waiting times in Radiation Oncology using machine learning

DOE Office of Scientific and Technical Information (OSTI.GOV)

Joseph, Ackeem; Herrera, David; Hijal, Tarek

We describe a method for predicting waiting times in radiation oncology. Machine learning is a powerful predictive modelling tool that benefits from large, potentially complex, datasets. The essence of machine learning is to predict future outcomes by learning from previous experience. The patient waiting experience remains one of the most vexing challenges facing healthcare. Waiting time uncertainty can cause patients, who are already sick and in pain, to worry about when they will receive the care they need. In radiation oncology, patients typically experience three types of waiting: Waiting at home for their treatment plan to be prepared Waiting inmore » the waiting room for daily radiotherapy Waiting in the waiting room to see a physician in consultation or follow-up These waiting periods are difficult for staff to predict and only rough estimates are typically provided, based on personal experience. In the present era of electronic health records, waiting times need not be so uncertain. At our centre, we have incorporated the electronic treatment records of all previously-treated patients into our machine learning model. We found that the Random Forest Regression model provides the best predictions for daily radiotherapy treatment waiting times (type 2). Using this model, we achieved a median residual (actual minus predicted value) of 0.25 minutes and a standard deviation residual of 6.5 minutes. The main features that generated the best fit model (from most to least significant) are: Allocated time, median past duration, fraction number and the number of treatment fields.« less

Segmentation of white matter hyperintensities using convolutional neural networks with global spatial information in routine clinical brain MRI with none or mild vascular pathology.

PubMed

Rachmadi, Muhammad Febrian; Valdés-Hernández, Maria Del C; Agan, Maria Leonora Fatimah; Di Perri, Carol; Komura, Taku

2018-06-01

We propose an adaptation of a convolutional neural network (CNN) scheme proposed for segmenting brain lesions with considerable mass-effect, to segment white matter hyperintensities (WMH) characteristic of brains with none or mild vascular pathology in routine clinical brain magnetic resonance images (MRI). This is a rather difficult segmentation problem because of the small area (i.e., volume) of the WMH and their similarity to non-pathological brain tissue. We investigate the effectiveness of the 2D CNN scheme by comparing its performance against those obtained from another deep learning approach: Deep Boltzmann Machine (DBM), two conventional machine learning approaches: Support Vector Machine (SVM) and Random Forest (RF), and a public toolbox: Lesion Segmentation Tool (LST), all reported to be useful for segmenting WMH in MRI. We also introduce a way to incorporate spatial information in convolution level of CNN for WMH segmentation named global spatial information (GSI). Analysis of covariance corroborated known associations between WMH progression, as assessed by all methods evaluated, and demographic and clinical data. Deep learning algorithms outperform conventional machine learning algorithms by excluding MRI artefacts and pathologies that appear similar to WMH. Our proposed approach of incorporating GSI also successfully helped CNN to achieve better automatic WMH segmentation regardless of network's settings tested. The mean Dice Similarity Coefficient (DSC) values for LST-LGA, SVM, RF, DBM, CNN and CNN-GSI were 0.2963, 0.1194, 0.1633, 0.3264, 0.5359 and 5389 respectively. Crown Copyright © 2018. Published by Elsevier Ltd. All rights reserved.

Real-data comparison of data mining methods in prediction of diabetes in iran.

PubMed

Tapak, Lily; Mahjub, Hossein; Hamidi, Omid; Poorolajal, Jalal

2013-09-01

Diabetes is one of the most common non-communicable diseases in developing countries. Early screening and diagnosis play an important role in effective prevention strategies. This study compared two traditional classification methods (logistic regression and Fisher linear discriminant analysis) and four machine-learning classifiers (neural networks, support vector machines, fuzzy c-mean, and random forests) to classify persons with and without diabetes. The data set used in this study included 6,500 subjects from the Iranian national non-communicable diseases risk factors surveillance obtained through a cross-sectional survey. The obtained sample was based on cluster sampling of the Iran population which was conducted in 2005-2009 to assess the prevalence of major non-communicable disease risk factors. Ten risk factors that are commonly associated with diabetes were selected to compare the performance of six classifiers in terms of sensitivity, specificity, total accuracy, and area under the receiver operating characteristic (ROC) curve criteria. Support vector machines showed the highest total accuracy (0.986) as well as area under the ROC (0.979). Also, this method showed high specificity (1.000) and sensitivity (0.820). All other methods produced total accuracy of more than 85%, but for all methods, the sensitivity values were very low (less than 0.350). The results of this study indicate that, in terms of sensitivity, specificity, and overall classification accuracy, the support vector machine model ranks first among all the classifiers tested in the prediction of diabetes. Therefore, this approach is a promising classifier for predicting diabetes, and it should be further investigated for the prediction of other diseases.

Can Statistical Machine Learning Algorithms Help for Classification of Obstructive Sleep Apnea Severity to Optimal Utilization of Polysomnography Resources?

PubMed

Bozkurt, Selen; Bostanci, Asli; Turhan, Murat

2017-08-11

The goal of this study is to evaluate the results of machine learning methods for the classification of OSA severity of patients with suspected sleep disorder breathing as normal, mild, moderate and severe based on non-polysomnographic variables: 1) clinical data, 2) symptoms and 3) physical examination. In order to produce classification models for OSA severity, five different machine learning methods (Bayesian network, Decision Tree, Random Forest, Neural Networks and Logistic Regression) were trained while relevant variables and their relationships were derived empirically from observed data. Each model was trained and evaluated using 10-fold cross-validation and to evaluate classification performances of all methods, true positive rate (TPR), false positive rate (FPR), Positive Predictive Value (PPV), F measure and Area Under Receiver Operating Characteristics curve (ROC-AUC) were used. Results of 10-fold cross validated tests with different variable settings promisingly indicated that the OSA severity of suspected OSA patients can be classified, using non-polysomnographic features, with 0.71 true positive rate as the highest and, 0.15 false positive rate as the lowest, respectively. Moreover, the test results of different variables settings revealed that the accuracy of the classification models was significantly improved when physical examination variables were added to the model. Study results showed that machine learning methods can be used to estimate the probabilities of no, mild, moderate, and severe obstructive sleep apnea and such approaches may improve accurate initial OSA screening and help referring only the suspected moderate or severe OSA patients to sleep laboratories for the expensive tests.

Reducing Sweeping Frequencies in Microwave NDT Employing Machine Learning Feature Selection

PubMed Central

Moomen, Abdelniser; Ali, Abdulbaset; Ramahi, Omar M.

2016-01-01

Nondestructive Testing (NDT) assessment of materials’ health condition is useful for classifying healthy from unhealthy structures or detecting flaws in metallic or dielectric structures. Performing structural health testing for coated/uncoated metallic or dielectric materials with the same testing equipment requires a testing method that can work on metallics and dielectrics such as microwave testing. Reducing complexity and expenses associated with current diagnostic practices of microwave NDT of structural health requires an effective and intelligent approach based on feature selection and classification techniques of machine learning. Current microwave NDT methods in general based on measuring variation in the S-matrix over the entire operating frequency ranges of the sensors. For instance, assessing the health of metallic structures using a microwave sensor depends on the reflection or/and transmission coefficient measurements as a function of the sweeping frequencies of the operating band. The aim of this work is reducing sweeping frequencies using machine learning feature selection techniques. By treating sweeping frequencies as features, the number of top important features can be identified, then only the most influential features (frequencies) are considered when building the microwave NDT equipment. The proposed method of reducing sweeping frequencies was validated experimentally using a waveguide sensor and a metallic plate with different cracks. Among the investigated feature selection techniques are information gain, gain ratio, relief, chi-squared. The effectiveness of the selected features were validated through performance evaluations of various classification models; namely, Nearest Neighbor, Neural Networks, Random Forest, and Support Vector Machine. Results showed good crack classification accuracy rates after employing feature selection algorithms. PMID:27104533

Multiple machine learning based descriptive and predictive workflow for the identification of potential PTP1B inhibitors.

PubMed

Chandra, Sharat; Pandey, Jyotsana; Tamrakar, Akhilesh Kumar; Siddiqi, Mohammad Imran

2017-01-01

In insulin and leptin signaling pathway, Protein-Tyrosine Phosphatase 1B (PTP1B) plays a crucial controlling role as a negative regulator, which makes it an attractive therapeutic target for both Type-2 Diabetes (T2D) and obesity. In this work, we have generated classification models by using the inhibition data set of known PTP1B inhibitors to identify new inhibitors of PTP1B utilizing multiple machine learning techniques like naïve Bayesian, random forest, support vector machine and k-nearest neighbors, along with structural fingerprints and selected molecular descriptors. Several models from each algorithm have been constructed and optimized, with the different combination of molecular descriptors and structural fingerprints. For the training and test sets, most of the predictive models showed more than 90% of overall prediction accuracies. The best model was obtained with support vector machine approach and has Matthews Correlation Coefficient of 0.82 for the external test set, which was further employed for the virtual screening of Maybridge small compound database. Five compounds were subsequently selected for experimental assay. Out of these two compounds were found to inhibit PTP1B with significant inhibitory activity in in-vitro inhibition assay. The structural fragments which are important for PTP1B inhibition were identified by naïve Bayesian method and can be further exploited to design new molecules around the identified scaffolds. The descriptive and predictive modeling strategy applied in this study is capable of identifying PTP1B inhibitors from the large compound libraries. Copyright © 2016 Elsevier Inc. All rights reserved.

Mapping the spatial distribution of Aedes aegypti and Aedes albopictus.

PubMed

Ding, Fangyu; Fu, Jingying; Jiang, Dong; Hao, Mengmeng; Lin, Gang

2018-02-01

Mosquito-borne infectious diseases, such as Rift Valley fever, Dengue, Chikungunya and Zika, have caused mass human death with the transnational expansion fueled by economic globalization. Simulating the distribution of the disease vectors is of great importance in formulating public health planning and disease control strategies. In the present study, we simulated the global distribution of Aedes aegypti and Aedes albopictus at a 5×5km spatial resolution with high-dimensional multidisciplinary datasets and machine learning methods Three relatively popular and robust machine learning models, including support vector machine (SVM), gradient boosting machine (GBM) and random forest (RF), were used. During the fine-tuning process based on training datasets of A. aegypti and A. albopictus, RF models achieved the highest performance with an area under the curve (AUC) of 0.973 and 0.974, respectively, followed by GBM (AUC of 0.971 and 0.972, respectively) and SVM (AUC of 0.963 and 0.964, respectively) models. The simulation difference between RF and GBM models was not statistically significant (p>0.05) based on the validation datasets, whereas statistically significant differences (p<0.05) were observed for RF and GBM simulations compared with SVM simulations. From the simulated maps derived from RF models, we observed that the distribution of A. albopictus was wider than that of A. aegypti along a latitudinal gradient. The discriminatory power of each factor in simulating the global distribution of the two species was also analyzed. Our results provided fundamental information for further study on disease transmission simulation and risk assessment. Copyright © 2017 Elsevier B.V. All rights reserved.

Implementation of Cyber-Physical Production Systems for Quality Prediction and Operation Control in Metal Casting.

PubMed

Lee, JuneHyuck; Noh, Sang Do; Kim, Hyun-Jung; Kang, Yong-Shin

2018-05-04

The prediction of internal defects of metal casting immediately after the casting process saves unnecessary time and money by reducing the amount of inputs into the next stage, such as the machining process, and enables flexible scheduling. Cyber-physical production systems (CPPS) perfectly fulfill the aforementioned requirements. This study deals with the implementation of CPPS in a real factory to predict the quality of metal casting and operation control. First, a CPPS architecture framework for quality prediction and operation control in metal-casting production was designed. The framework describes collaboration among internet of things (IoT), artificial intelligence, simulations, manufacturing execution systems, and advanced planning and scheduling systems. Subsequently, the implementation of the CPPS in actual plants is described. Temperature is a major factor that affects casting quality, and thus, temperature sensors and IoT communication devices were attached to casting machines. The well-known NoSQL database, HBase and the high-speed processing/analysis tool, Spark, are used for IoT repository and data pre-processing, respectively. Many machine learning algorithms such as decision tree, random forest, artificial neural network, and support vector machine were used for quality prediction and compared with R software. Finally, the operation of the entire system is demonstrated through a CPPS dashboard. In an era in which most CPPS-related studies are conducted on high-level abstract models, this study describes more specific architectural frameworks, use cases, usable software, and analytical methodologies. In addition, this study verifies the usefulness of CPPS by estimating quantitative effects. This is expected to contribute to the proliferation of CPPS in the industry.

Improving virtual screening predictive accuracy of Human kallikrein 5 inhibitors using machine learning models.

PubMed

Fang, Xingang; Bagui, Sikha; Bagui, Subhash

2017-08-01

The readily available high throughput screening (HTS) data from the PubChem database provides an opportunity for mining of small molecules in a variety of biological systems using machine learning techniques. From the thousands of available molecular descriptors developed to encode useful chemical information representing the characteristics of molecules, descriptor selection is an essential step in building an optimal quantitative structural-activity relationship (QSAR) model. For the development of a systematic descriptor selection strategy, we need the understanding of the relationship between: (i) the descriptor selection; (ii) the choice of the machine learning model; and (iii) the characteristics of the target bio-molecule. In this work, we employed the Signature descriptor to generate a dataset on the Human kallikrein 5 (hK 5) inhibition confirmatory assay data and compared multiple classification models including logistic regression, support vector machine, random forest and k-nearest neighbor. Under optimal conditions, the logistic regression model provided extremely high overall accuracy (98%) and precision (90%), with good sensitivity (65%) in the cross validation test. In testing the primary HTS screening data with more than 200K molecular structures, the logistic regression model exhibited the capability of eliminating more than 99.9% of the inactive structures. As part of our exploration of the descriptor-model-target relationship, the excellent predictive performance of the combination of the Signature descriptor and the logistic regression model on the assay data of the Human kallikrein 5 (hK 5) target suggested a feasible descriptor/model selection strategy on similar targets. Copyright © 2017 Elsevier Ltd. All rights reserved.

Enhancement of plant metabolite fingerprinting by machine learning.

PubMed

Scott, Ian M; Vermeer, Cornelia P; Liakata, Maria; Corol, Delia I; Ward, Jane L; Lin, Wanchang; Johnson, Helen E; Whitehead, Lynne; Kular, Baldeep; Baker, John M; Walsh, Sean; Dave, Anuja; Larson, Tony R; Graham, Ian A; Wang, Trevor L; King, Ross D; Draper, John; Beale, Michael H

2010-08-01

Metabolite fingerprinting of Arabidopsis (Arabidopsis thaliana) mutants with known or predicted metabolic lesions was performed by (1)H-nuclear magnetic resonance, Fourier transform infrared, and flow injection electrospray-mass spectrometry. Fingerprinting enabled processing of five times more plants than conventional chromatographic profiling and was competitive for discriminating mutants, other than those affected in only low-abundance metabolites. Despite their rapidity and complexity, fingerprints yielded metabolomic insights (e.g. that effects of single lesions were usually not confined to individual pathways). Among fingerprint techniques, (1)H-nuclear magnetic resonance discriminated the most mutant phenotypes from the wild type and Fourier transform infrared discriminated the fewest. To maximize information from fingerprints, data analysis was crucial. One-third of distinctive phenotypes might have been overlooked had data models been confined to principal component analysis score plots. Among several methods tested, machine learning (ML) algorithms, namely support vector machine or random forest (RF) classifiers, were unsurpassed for phenotype discrimination. Support vector machines were often the best performing classifiers, but RFs yielded some particularly informative measures. First, RFs estimated margins between mutant phenotypes, whose relations could then be visualized by Sammon mapping or hierarchical clustering. Second, RFs provided importance scores for the features within fingerprints that discriminated mutants. These scores correlated with analysis of variance F values (as did Kruskal-Wallis tests, true- and false-positive measures, mutual information, and the Relief feature selection algorithm). ML classifiers, as models trained on one data set to predict another, were ideal for focused metabolomic queries, such as the distinctiveness and consistency of mutant phenotypes. Accessible software for use of ML in plant physiology is highlighted.

Mapping Forest Height in Gabon Using UAVSAR Multi-Baseline Polarimetric SAR Interferometry and Lidar Fusion

NASA Astrophysics Data System (ADS)

Simard, M.; Denbina, M. W.

2017-12-01

Using data collected by NASA's Uninhabited Aerial Vehicle Synthetic Aperture Radar (UAVSAR) and Land, Vegetation, and Ice Sensor (LVIS) lidar, we have estimated forest canopy height for a number of study areas in the country of Gabon using a new machine learning data fusion approach. Using multi-baseline polarimetric synthetic aperture radar interferometry (PolInSAR) data collected by UAVSAR, forest heights can be estimated using the random volume over ground model. In the case of multi-baseline UAVSAR data consisting of many repeat passes with spatially separated flight tracks, we can estimate different forest height values for each different image pair, or baseline. In order to choose the best forest height estimate for each pixel, the baselines must be selected or ranked, taking care to avoid baselines with unsuitable spatial separation, or severe temporal decorrelation effects. The current baseline selection algorithms in the literature use basic quality metrics derived from the PolInSAR data which are not necessarily indicative of the true height accuracy in all cases. We have developed a new data fusion technique which treats PolInSAR baseline selection as a supervised classification problem, where the classifier is trained using a sparse sampling of lidar data within the PolInSAR coverage area. The classifier uses a large variety of PolInSAR-derived features as input, including radar backscatter as well as features based on the PolInSAR coherence region shape and the PolInSAR complex coherences. The resulting data fusion method produces forest height estimates which are more accurate than a purely radar-based approach, while having a larger coverage area than the input lidar training data, combining some of the strengths of each sensor. The technique demonstrates the strong potential for forest canopy height and above-ground biomass mapping using fusion of PolInSAR with data from future spaceborne lidar missions such as the upcoming Global Ecosystems Dynamics Investigation (GEDI) lidar.

Comparison of machine learning techniques to predict all-cause mortality using fitness data: the Henry ford exercIse testing (FIT) project.

PubMed

Sakr, Sherif; Elshawi, Radwa; Ahmed, Amjad M; Qureshi, Waqas T; Brawner, Clinton A; Keteyian, Steven J; Blaha, Michael J; Al-Mallah, Mouaz H

2017-12-19

Prior studies have demonstrated that cardiorespiratory fitness (CRF) is a strong marker of cardiovascular health. Machine learning (ML) can enhance the prediction of outcomes through classification techniques that classify the data into predetermined categories. The aim of this study is to present an evaluation and comparison of how machine learning techniques can be applied on medical records of cardiorespiratory fitness and how the various techniques differ in terms of capabilities of predicting medical outcomes (e.g. mortality). We use data of 34,212 patients free of known coronary artery disease or heart failure who underwent clinician-referred exercise treadmill stress testing at Henry Ford Health Systems Between 1991 and 2009 and had a complete 10-year follow-up. Seven machine learning classification techniques were evaluated: Decision Tree (DT), Support Vector Machine (SVM), Artificial Neural Networks (ANN), Naïve Bayesian Classifier (BC), Bayesian Network (BN), K-Nearest Neighbor (KNN) and Random Forest (RF). In order to handle the imbalanced dataset used, the Synthetic Minority Over-Sampling Technique (SMOTE) is used. Two set of experiments have been conducted with and without the SMOTE sampling technique. On average over different evaluation metrics, SVM Classifier has shown the lowest performance while other models like BN, BC and DT performed better. The RF classifier has shown the best performance (AUC = 0.97) among all models trained using the SMOTE sampling. The results show that various ML techniques can significantly vary in terms of its performance for the different evaluation metrics. It is also not necessarily that the more complex the ML model, the more prediction accuracy can be achieved. The prediction performance of all models trained with SMOTE is much better than the performance of models trained without SMOTE. The study shows the potential of machine learning methods for predicting all-cause mortality using cardiorespiratory fitness data.

Using Random Forest Models to Predict Organizational Violence

NASA Technical Reports Server (NTRS)

Levine, Burton; Bobashev, Georgly

2012-01-01

We present a methodology to access the proclivity of an organization to commit violence against nongovernment personnel. We fitted a Random Forest model using the Minority at Risk Organizational Behavior (MAROS) dataset. The MAROS data is longitudinal; so, individual observations are not independent. We propose a modification to the standard Random Forest methodology to account for the violation of the independence assumption. We present the results of the model fit, an example of predicting violence for an organization; and finally, we present a summary of the forest in a "meta-tree,"

Comparison of Nine Statistical Model Based Warfarin Pharmacogenetic Dosing Algorithms Using the Racially Diverse International Warfarin Pharmacogenetic Consortium Cohort Database

PubMed Central

Liu, Rong; Li, Xi; Zhang, Wei; Zhou, Hong-Hao

2015-01-01

Objective Multiple linear regression (MLR) and machine learning techniques in pharmacogenetic algorithm-based warfarin dosing have been reported. However, performances of these algorithms in racially diverse group have never been objectively evaluated and compared. In this literature-based study, we compared the performances of eight machine learning techniques with those of MLR in a large, racially-diverse cohort. Methods MLR, artificial neural network (ANN), regression tree (RT), multivariate adaptive regression splines (MARS), boosted regression tree (BRT), support vector regression (SVR), random forest regression (RFR), lasso regression (LAR) and Bayesian additive regression trees (BART) were applied in warfarin dose algorithms in a cohort from the International Warfarin Pharmacogenetics Consortium database. Covariates obtained by stepwise regression from 80% of randomly selected patients were used to develop algorithms. To compare the performances of these algorithms, the mean percentage of patients whose predicted dose fell within 20% of the actual dose (mean percentage within 20%) and the mean absolute error (MAE) were calculated in the remaining 20% of patients. The performances of these techniques in different races, as well as the dose ranges of therapeutic warfarin were compared. Robust results were obtained after 100 rounds of resampling. Results BART, MARS and SVR were statistically indistinguishable and significantly out performed all the other approaches in the whole cohort (MAE: 8.84–8.96 mg/week, mean percentage within 20%: 45.88%–46.35%). In the White population, MARS and BART showed higher mean percentage within 20% and lower mean MAE than those of MLR (all p values < 0.05). In the Asian population, SVR, BART, MARS and LAR performed the same as MLR. MLR and LAR optimally performed among the Black population. When patients were grouped in terms of warfarin dose range, all machine learning techniques except ANN and LAR showed significantly higher mean percentage within 20%, and lower MAE (all p values < 0.05) than MLR in the low- and high- dose ranges. Conclusion Overall, machine learning-based techniques, BART, MARS and SVR performed superior than MLR in warfarin pharmacogenetic dosing. Differences of algorithms’ performances exist among the races. Moreover, machine learning-based algorithms tended to perform better in the low- and high- dose ranges than MLR. PMID:26305568

Lumber Scanning System for Surface Defect Detection

Treesearch

D. Earl Kline; Y. Jason Hou; Richard W. Conners; Daniel L. Schmoldt; Philip A. Araman

1992-01-01

This paper describes research aimed at developing a machine vision technology to drive automated processes in the hardwood forest products manufacturing industry. An industrial-scale machine vision system has been designed to scan variable-size hardwood lumber for detecting important features that influence the grade and value of lumber such as knots, holes, wane,...

Machine Vision Systems for Processing Hardwood Lumber and Logs

Treesearch

Philip A. Araman; Daniel L. Schmoldt; Tai-Hoon Cho; Dongping Zhu; Richard W. Conners; D. Earl Kline

1992-01-01

Machine vision and automated processing systems are under development at Virginia Tech University with support and cooperation from the USDA Forest Service. Our goals are to help U.S. hardwood producers automate, reduce costs, increase product volume and value recovery, and market higher value, more accurately graded and described products. Any vision system is...

Robust Spatial Autoregressive Modeling for Hardwood Log Inspection

Treesearch

Dongping Zhu; A.A. Beex

1994-01-01

We explore the application of a stochastic texture modeling method toward a machine vision system for log inspection in the forest products industry. This machine vision system uses computerized tomography (CT) imaging to locate and identify internal defects in hardwood logs. The application of CT to such industrial vision problems requires efficient and robust image...

Automatic machine-learning based identification of jogging periods from accelerometer measurements of adolescents under field conditions.

PubMed

Zdravevski, Eftim; Risteska Stojkoska, Biljana; Standl, Marie; Schulz, Holger

2017-01-01

Assessment of health benefits associated with physical activity depend on the activity duration, intensity and frequency, therefore their correct identification is very valuable and important in epidemiological and clinical studies. The aims of this study are: to develop an algorithm for automatic identification of intended jogging periods; and to assess whether the identification performance is improved when using two accelerometers at the hip and ankle, compared to when using only one at either position. The study used diarized jogging periods and the corresponding accelerometer data from thirty-nine, 15-year-old adolescents, collected under field conditions, as part of the GINIplus study. The data was obtained from two accelerometers placed at the hip and ankle. Automated feature engineering technique was performed to extract features from the raw accelerometer readings and to select a subset of the most significant features. Four machine learning algorithms were used for classification: Logistic regression, Support Vector Machines, Random Forest and Extremely Randomized Trees. Classification was performed using only data from the hip accelerometer, using only data from ankle accelerometer and using data from both accelerometers. The reported jogging periods were verified by visual inspection and used as golden standard. After the feature selection and tuning of the classification algorithms, all options provided a classification accuracy of at least 0.99, independent of the applied segmentation strategy with sliding windows of either 60s or 180s. The best matching ratio, i.e. the length of correctly identified jogging periods related to the total time including the missed ones, was up to 0.875. It could be additionally improved up to 0.967 by application of post-classification rules, which considered the duration of breaks and jogging periods. There was no obvious benefit of using two accelerometers, rather almost the same performance could be achieved from either accelerometer position. Machine learning techniques can be used for automatic activity recognition, as they provide very accurate activity recognition, significantly more accurate than when keeping a diary. Identification of jogging periods in adolescents can be performed using only one accelerometer. Performance-wise there is no significant benefit from using accelerometers on both locations.

Machine-learned analysis of the association of next-generation sequencing-based human TRPV1 and TRPA1 genotypes with the sensitivity to heat stimuli and topically applied capsaicin.

PubMed

Kringel, Dario; Geisslinger, Gerd; Resch, Eduard; Oertel, Bruno G; Thrun, Michael C; Heinemann, Sarah; Lötsch, Jörn

2018-03-27

Heat pain and its modulation by capsaicin varies among subjects in experimental and clinical settings. A plausible cause is a genetic component, of which TRPV1 ion channels, by their response to both heat and capsaicin, are primary candidates. However, TRPA1 channels can heterodimerize with TRPV1 channels and carry genetic variants reported to modulate heat pain sensitivity. To address the role of these candidate genes in capsaicin-induced hypersensitization to heat, pain thresholds acquired before and after topical application of capsaicin and TRPA1/TRPV1 exomic sequences derived by next-generation sequencing were assessed in n = 75 healthy volunteers and the genetic information comprised 278 loci. Gaussian mixture modeling indicated 2 phenotype groups with high or low capsaicin-induced hypersensitization to heat. Unsupervised machine learning implemented as swarm-based clustering hinted at differences in the genetic pattern between these phenotype groups. Several methods of supervised machine learning implemented as random forests, adaptive boosting, k-nearest neighbors, naive Bayes, support vector machines, and for comparison, binary logistic regression predicted the phenotype group association consistently better when based on the observed genotypes than when using a random permutation of the exomic sequences. Of note, TRPA1 variants were more important for correct phenotype group association than TRPV1 variants. This indicates a role of the TRPA1 and TRPV1 next-generation sequencing-based genetic pattern in the modulation of the individual response to heat-related pain phenotypes. When considering earlier evidence that topical capsaicin can induce neuropathy-like quantitative sensory testing patterns in healthy subjects, implications for future analgesic treatments with transient receptor potential inhibitors arise.This is an open-access article distributed under the terms of the Creative Commons Attribution-Non Commercial-No Derivatives License 4.0 (CCBY-NC-ND), where it is permissible to download and share the work provided it is properly cited. The work cannot be changed in any way or used commercially without permission from the journal.

A statistical assessment of population trends for data deficient Mexican amphibians

PubMed Central

Thessen, Anne E.; Arias-Caballero, Paulina; Ayala-Orozco, Bárbara

2014-01-01

Background. Mexico has the world’s fifth largest population of amphibians and the second country with the highest quantity of threatened amphibian species. About 10% of Mexican amphibians lack enough data to be assigned to a risk category by the IUCN, so in this paper we want to test a statistical tool that, in the absence of specific demographic data, can assess a species’ risk of extinction, population trend, and to better understand which variables increase their vulnerability. Recent studies have demonstrated that the risk of species decline depends on extrinsic and intrinsic traits, thus including both of them for assessing extinction might render more accurate assessment of threats. Methods. We harvested data from the Encyclopedia of Life (EOL) and the published literature for Mexican amphibians, and used these data to assess the population trend of some of the Mexican species that have been assigned to the Data Deficient category of the IUCN using Random Forests, a Machine Learning method that gives a prediction of complex processes and identifies the most important variables that account for the predictions. Results. Our results show that most of the data deficient Mexican amphibians that we used have decreasing population trends. We found that Random Forests is a solid way to identify species with decreasing population trends when no demographic data is available. Moreover, we point to the most important variables that make species more vulnerable for extinction. This exercise is a very valuable first step in assigning conservation priorities for poorly known species. PMID:25548736

Machine learning based brain tumour segmentation on limited data using local texture and abnormality.

PubMed

Bonte, Stijn; Goethals, Ingeborg; Van Holen, Roel

2018-05-07

Brain tumour segmentation in medical images is a very challenging task due to the large variety in tumour shape, position, appearance, scanning modalities and scanning parameters. Most existing segmentation algorithms use information from four different MRI-sequences, but since this is often not available, there is need for a method able to delineate the different tumour tissues based on a minimal amount of data. We present a novel approach using a Random Forests model combining voxelwise texture and abnormality features on a contrast-enhanced T1 and FLAIR MRI. We transform the two scans into 275 feature maps. A random forest model next calculates the probability to belong to 4 tumour classes or 5 normal classes. Afterwards, a dedicated voxel clustering algorithm provides the final tumour segmentation. We trained our method on the BraTS 2013 database and validated it on the larger BraTS 2017 dataset. We achieve median Dice scores of 40.9% (low-grade glioma) and 75.0% (high-grade glioma) to delineate the active tumour, and 68.4%/80.1% for the total abnormal region including edema. Our fully automated brain tumour segmentation algorithm is able to delineate contrast enhancing tissue and oedema with high accuracy based only on post-contrast T1-weighted and FLAIR MRI, whereas for non-enhancing tumour tissue and necrosis only moderate results are obtained. This makes the method especially suitable for high-grade glioma. Copyright © 2018 Elsevier Ltd. All rights reserved.

Predicting Long-Term Cognitive Outcome Following Breast Cancer with Pre-Treatment Resting State fMRI and Random Forest Machine Learning.

PubMed

Kesler, Shelli R; Rao, Arvind; Blayney, Douglas W; Oakley-Girvan, Ingrid A; Karuturi, Meghan; Palesh, Oxana

2017-01-01

We aimed to determine if resting state functional magnetic resonance imaging (fMRI) acquired at pre-treatment baseline could accurately predict breast cancer-related cognitive impairment at long-term follow-up. We evaluated 31 patients with breast cancer (age 34-65) prior to any treatment, post-chemotherapy and 1 year later. Cognitive testing scores were normalized based on data obtained from 43 healthy female controls and then used to categorize patients as impaired or not based on longitudinal changes. We measured clustering coefficient, a measure of local connectivity, by applying graph theory to baseline resting state fMRI and entered these metrics along with relevant patient-related and medical variables into random forest classification. Incidence of cognitive impairment at 1 year follow-up was 55% and was predicted by classification algorithms with up to 100% accuracy ( p < 0.0001). The neuroimaging-based model was significantly more accurate than a model involving patient-related and medical variables ( p = 0.005). Hub regions belonging to several distinct functional networks were the most important predictors of cognitive outcome. Characteristics of these hubs indicated potential spread of brain injury from default mode to other networks over time. These findings suggest that resting state fMRI is a promising tool for predicting future cognitive impairment associated with breast cancer. This information could inform treatment decision making by identifying patients at highest risk for long-term cognitive impairment.

Predicting Long-Term Cognitive Outcome Following Breast Cancer with Pre-Treatment Resting State fMRI and Random Forest Machine Learning

PubMed Central

Kesler, Shelli R.; Rao, Arvind; Blayney, Douglas W.; Oakley-Girvan, Ingrid A.; Karuturi, Meghan; Palesh, Oxana

2017-01-01

We aimed to determine if resting state functional magnetic resonance imaging (fMRI) acquired at pre-treatment baseline could accurately predict breast cancer-related cognitive impairment at long-term follow-up. We evaluated 31 patients with breast cancer (age 34–65) prior to any treatment, post-chemotherapy and 1 year later. Cognitive testing scores were normalized based on data obtained from 43 healthy female controls and then used to categorize patients as impaired or not based on longitudinal changes. We measured clustering coefficient, a measure of local connectivity, by applying graph theory to baseline resting state fMRI and entered these metrics along with relevant patient-related and medical variables into random forest classification. Incidence of cognitive impairment at 1 year follow-up was 55% and was predicted by classification algorithms with up to 100% accuracy (p < 0.0001). The neuroimaging-based model was significantly more accurate than a model involving patient-related and medical variables (p = 0.005). Hub regions belonging to several distinct functional networks were the most important predictors of cognitive outcome. Characteristics of these hubs indicated potential spread of brain injury from default mode to other networks over time. These findings suggest that resting state fMRI is a promising tool for predicting future cognitive impairment associated with breast cancer. This information could inform treatment decision making by identifying patients at highest risk for long-term cognitive impairment. PMID:29187817

Neural Predictors of Initiating Alcohol Use During Adolescence.

PubMed

Squeglia, Lindsay M; Ball, Tali M; Jacobus, Joanna; Brumback, Ty; McKenna, Benjamin S; Nguyen-Louie, Tam T; Sorg, Scott F; Paulus, Martin P; Tapert, Susan F

2017-02-01

Underage drinking is widely recognized as a leading public health and social problem for adolescents in the United States. Being able to identify at-risk adolescents before they initiate heavy alcohol use could have important clinical and public health implications; however, few investigations have explored individual-level precursors of adolescent substance use. This prospective investigation used machine learning with demographic, neurocognitive, and neuroimaging data in substance-naive adolescents to identify predictors of alcohol use initiation by age 18. Participants (N=137) were healthy substance-naive adolescents (ages 12-14) who underwent neuropsychological testing and structural and functional magnetic resonance imaging (sMRI and fMRI), and then were followed annually. By age 18, 70 youths (51%) initiated moderate to heavy alcohol use, and 67 remained nonusers. Random forest classification models identified the most important predictors of alcohol use from a large set of demographic, neuropsychological, sMRI, and fMRI variables. Random forest models identified 34 predictors contributing to alcohol use by age 18, including several demographic and behavioral factors (being male, higher socioeconomic status, early dating, more externalizing behaviors, positive alcohol expectancies), worse executive functioning, and thinner cortices and less brain activation in diffusely distributed regions of the brain. Incorporating a mix of demographic, behavioral, neuropsychological, and neuroimaging data may be the best strategy for identifying youths at risk for initiating alcohol use during adolescence. The identified risk factors will be useful for alcohol prevention efforts and in research to address brain mechanisms that may contribute to early drinking.

Random forest classification of large volume structures for visuo-haptic rendering in CT images

NASA Astrophysics Data System (ADS)

Mastmeyer, Andre; Fortmeier, Dirk; Handels, Heinz

2016-03-01

For patient-specific voxel-based visuo-haptic rendering of CT scans of the liver area, the fully automatic segmentation of large volume structures such as skin, soft tissue, lungs and intestine (risk structures) is important. Using a machine learning based approach, several existing segmentations from 10 segmented gold-standard patients are learned by random decision forests individually and collectively. The core of this paper is feature selection and the application of the learned classifiers to a new patient data set. In a leave-some-out cross-validation, the obtained full volume segmentations are compared to the gold-standard segmentations of the untrained patients. The proposed classifiers use a multi-dimensional feature space to estimate the hidden truth, instead of relying on clinical standard threshold and connectivity based methods. The result of our efficient whole-body section classification are multi-label maps with the considered tissues. For visuo-haptic simulation, other small volume structures would have to be segmented additionally. We also take a look into these structures (liver vessels). For an experimental leave-some-out study consisting of 10 patients, the proposed method performs much more efficiently compared to state of the art methods. In two variants of leave-some-out experiments we obtain best mean DICE ratios of 0.79, 0.97, 0.63 and 0.83 for skin, soft tissue, hard bone and risk structures. Liver structures are segmented with DICE 0.93 for the liver, 0.43 for blood vessels and 0.39 for bile vessels.

A statistical assessment of population trends for data deficient Mexican amphibians.

PubMed

Quintero, Esther; Thessen, Anne E; Arias-Caballero, Paulina; Ayala-Orozco, Bárbara

2014-01-01

Background. Mexico has the world's fifth largest population of amphibians and the second country with the highest quantity of threatened amphibian species. About 10% of Mexican amphibians lack enough data to be assigned to a risk category by the IUCN, so in this paper we want to test a statistical tool that, in the absence of specific demographic data, can assess a species' risk of extinction, population trend, and to better understand which variables increase their vulnerability. Recent studies have demonstrated that the risk of species decline depends on extrinsic and intrinsic traits, thus including both of them for assessing extinction might render more accurate assessment of threats. Methods. We harvested data from the Encyclopedia of Life (EOL) and the published literature for Mexican amphibians, and used these data to assess the population trend of some of the Mexican species that have been assigned to the Data Deficient category of the IUCN using Random Forests, a Machine Learning method that gives a prediction of complex processes and identifies the most important variables that account for the predictions. Results. Our results show that most of the data deficient Mexican amphibians that we used have decreasing population trends. We found that Random Forests is a solid way to identify species with decreasing population trends when no demographic data is available. Moreover, we point to the most important variables that make species more vulnerable for extinction. This exercise is a very valuable first step in assigning conservation priorities for poorly known species.

Local receptive field constrained stacked sparse autoencoder for classification of hyperspectral images.

PubMed

Wan, Xiaoqing; Zhao, Chunhui

2017-06-01

As a competitive machine learning algorithm, the stacked sparse autoencoder (SSA) has achieved outstanding popularity in exploiting high-level features for classification of hyperspectral images (HSIs). In general, in the SSA architecture, the nodes between adjacent layers are fully connected and need to be iteratively fine-tuned during the pretraining stage; however, the nodes of previous layers further away may be less likely to have a dense correlation to the given node of subsequent layers. Therefore, to reduce the classification error and increase the learning rate, this paper proposes the general framework of locally connected SSA; that is, the biologically inspired local receptive field (LRF) constrained SSA architecture is employed to simultaneously characterize the local correlations of spectral features and extract high-level feature representations of hyperspectral data. In addition, the appropriate receptive field constraint is concurrently updated by measuring the spatial distances from the neighbor nodes to the corresponding node. Finally, the efficient random forest classifier is cascaded to the last hidden layer of the SSA architecture as a benchmark classifier. Experimental results on two real HSI datasets demonstrate that the proposed hierarchical LRF constrained stacked sparse autoencoder and random forest (SSARF) provides encouraging results with respect to other contrastive methods, for instance, the improvements of overall accuracy in a range of 0.72%-10.87% for the Indian Pines dataset and 0.74%-7.90% for the Kennedy Space Center dataset; moreover, it generates lower running time compared with the result provided by similar SSARF based methodology.

CNV-RF Is a Random Forest-Based Copy Number Variation Detection Method Using Next-Generation Sequencing.

PubMed

Onsongo, Getiria; Baughn, Linda B; Bower, Matthew; Henzler, Christine; Schomaker, Matthew; Silverstein, Kevin A T; Thyagarajan, Bharat

2016-11-01

Simultaneous detection of small copy number variations (CNVs) (<0.5 kb) and single-nucleotide variants in clinically significant genes is of great interest for clinical laboratories. The analytical variability in next-generation sequencing (NGS) and artifacts in coverage data because of issues with mappability along with lack of robust bioinformatics tools for CNV detection have limited the utility of targeted NGS data to identify CNVs. We describe the development and implementation of a bioinformatics algorithm, copy number variation-random forest (CNV-RF), that incorporates a machine learning component to identify CNVs from targeted NGS data. Using CNV-RF, we identified 12 of 13 deletions in samples with known CNVs, two cases with duplications, and identified novel deletions in 22 additional cases. Furthermore, no CNVs were identified among 60 genes in 14 cases with normal copy number and no CNVs were identified in another 104 patients with clinical suspicion of CNVs. All positive deletions and duplications were confirmed using a quantitative PCR method. CNV-RF also detected heterozygous deletions and duplications with a specificity of 50% across 4813 genes. The ability of CNV-RF to detect clinically relevant CNVs with a high degree of sensitivity along with confirmation using a low-cost quantitative PCR method provides a framework for providing comprehensive NGS-based CNV/single-nucleotide variant detection in a clinical molecular diagnostics laboratory. Copyright © 2016 American Society for Investigative Pathology and the Association for Molecular Pathology. Published by Elsevier Inc. All rights reserved.

Differentiation of Crataegus spp. guided by nuclear magnetic resonance spectrometry with chemometric analyses.

PubMed

Lund, Jensen A; Brown, Paula N; Shipley, Paul R

2017-09-01

For compliance with US Current Good Manufacturing Practice regulations for dietary supplements, manufacturers must provide identity of source plant material. Despite the popularity of hawthorn as a dietary supplement, relatively little is known about the comparative phytochemistry of different hawthorn species, and in particular North American hawthorns. The combination of NMR spectrometry with chemometric analyses offers an innovative approach to differentiating hawthorn species and exploring the phytochemistry. Two European and two North American species, harvested from a farm trial in late summer 2008, were analyzed by standard 1D 1 H and J-resolved (JRES) experiments. The data were preprocessed and modelled by principal component analysis (PCA). A supervised model was then generated by partial least squares-discriminant analysis (PLS-DA) for classification and evaluated by cross validation. Supervised random forests models were constructed from the dataset to explore the potential of machine learning for identification of unique patterns across species. 1D 1 H NMR data yielded increased differentiation over the JRES data. The random forests results correlated with PLS-DA results and outperformed PLS-DA in classification accuracy. In all of these analyses differentiation of the Crataegus spp. was best achieved by focusing on the NMR spectral region that contains signals unique to plant phenolic compounds. Identification of potentially significant metabolites for differentiation between species was approached using univariate techniques including significance analysis of microarrays and Kruskall-Wallis tests. Copyright © 2017 Elsevier Ltd. All rights reserved.

A novel transferable individual tree crown delineation model based on Fishing Net Dragging and boundary classification

NASA Astrophysics Data System (ADS)

Liu, Tao; Im, Jungho; Quackenbush, Lindi J.

2015-12-01

This study provides a novel approach to individual tree crown delineation (ITCD) using airborne Light Detection and Ranging (LiDAR) data in dense natural forests using two main steps: crown boundary refinement based on a proposed Fishing Net Dragging (FiND) method, and segment merging based on boundary classification. FiND starts with approximate tree crown boundaries derived using a traditional watershed method with Gaussian filtering and refines these boundaries using an algorithm that mimics how a fisherman drags a fishing net. Random forest machine learning is then used to classify boundary segments into two classes: boundaries between trees and boundaries between branches that belong to a single tree. Three groups of LiDAR-derived features-two from the pseudo waveform generated along with crown boundaries and one from a canopy height model (CHM)-were used in the classification. The proposed ITCD approach was tested using LiDAR data collected over a mountainous region in the Adirondack Park, NY, USA. Overall accuracy of boundary classification was 82.4%. Features derived from the CHM were generally more important in the classification than the features extracted from the pseudo waveform. A comprehensive accuracy assessment scheme for ITCD was also introduced by considering both area of crown overlap and crown centroids. Accuracy assessment using this new scheme shows the proposed ITCD achieved 74% and 78% as overall accuracy, respectively, for deciduous and mixed forest.

Sequence Based Prediction of DNA-Binding Proteins Based on Hybrid Feature Selection Using Random Forest and Gaussian Naïve Bayes

PubMed Central

Lou, Wangchao; Wang, Xiaoqing; Chen, Fan; Chen, Yixiao; Jiang, Bo; Zhang, Hua

2014-01-01

Developing an efficient method for determination of the DNA-binding proteins, due to their vital roles in gene regulation, is becoming highly desired since it would be invaluable to advance our understanding of protein functions. In this study, we proposed a new method for the prediction of the DNA-binding proteins, by performing the feature rank using random forest and the wrapper-based feature selection using forward best-first search strategy. The features comprise information from primary sequence, predicted secondary structure, predicted relative solvent accessibility, and position specific scoring matrix. The proposed method, called DBPPred, used Gaussian naïve Bayes as the underlying classifier since it outperformed five other classifiers, including decision tree, logistic regression, k-nearest neighbor, support vector machine with polynomial kernel, and support vector machine with radial basis function. As a result, the proposed DBPPred yields the highest average accuracy of 0.791 and average MCC of 0.583 according to the five-fold cross validation with ten runs on the training benchmark dataset PDB594. Subsequently, blind tests on the independent dataset PDB186 by the proposed model trained on the entire PDB594 dataset and by other five existing methods (including iDNA-Prot, DNA-Prot, DNAbinder, DNABIND and DBD-Threader) were performed, resulting in that the proposed DBPPred yielded the highest accuracy of 0.769, MCC of 0.538, and AUC of 0.790. The independent tests performed by the proposed DBPPred on completely a large non-DNA binding protein dataset and two RNA binding protein datasets also showed improved or comparable quality when compared with the relevant prediction methods. Moreover, we observed that majority of the selected features by the proposed method are statistically significantly different between the mean feature values of the DNA-binding and the non DNA-binding proteins. All of the experimental results indicate that the proposed DBPPred can be an alternative perspective predictor for large-scale determination of DNA-binding proteins. PMID:24475169

An integrated method for atherosclerotic carotid plaque segmentation in ultrasound image.

PubMed

Qian, Chunjun; Yang, Xiaoping

2018-01-01

Carotid artery atherosclerosis is an important cause of stroke. Ultrasound imaging has been widely used in the diagnosis of atherosclerosis. Therefore, segmenting atherosclerotic carotid plaque in ultrasound image is an important task. Accurate plaque segmentation is helpful for the measurement of carotid plaque burden. In this paper, we propose and evaluate a novel learning-based integrated framework for plaque segmentation. In our study, four different classification algorithms, along with the auto-context iterative algorithm, were employed to effectively integrate features from ultrasound images and later also the iteratively estimated and refined probability maps together for pixel-wise classification. The four classification algorithms were support vector machine with linear kernel, support vector machine with radial basis function kernel, AdaBoost and random forest. The plaque segmentation was implemented in the generated probability map. The performance of the four different learning-based plaque segmentation methods was tested on 29 B-mode ultrasound images. The evaluation indices for our proposed methods were consisted of sensitivity, specificity, Dice similarity coefficient, overlap index, error of area, absolute error of area, point-to-point distance, and Hausdorff point-to-point distance, along with the area under the ROC curve. The segmentation method integrated the random forest and an auto-context model obtained the best results (sensitivity 80.4 ± 8.4%, specificity 96.5 ± 2.0%, Dice similarity coefficient 81.0 ± 4.1%, overlap index 68.3 ± 5.8%, error of area -1.02 ± 18.3%, absolute error of area 14.7 ± 10.9%, point-to-point distance 0.34 ± 0.10 mm, Hausdorff point-to-point distance 1.75 ± 1.02 mm, and area under the ROC curve 0.897), which were almost the best, compared with that from the existed methods. Our proposed learning-based integrated framework investigated in this study could be useful for atherosclerotic carotid plaque segmentation, which will be helpful for the measurement of carotid plaque burden. Copyright © 2017 Elsevier B.V. All rights reserved.

ReliefSeq: A Gene-Wise Adaptive-K Nearest-Neighbor Feature Selection Tool for Finding Gene-Gene Interactions and Main Effects in mRNA-Seq Gene Expression Data

PubMed Central

McKinney, Brett A.; White, Bill C.; Grill, Diane E.; Li, Peter W.; Kennedy, Richard B.; Poland, Gregory A.; Oberg, Ann L.

2013-01-01

Relief-F is a nonparametric, nearest-neighbor machine learning method that has been successfully used to identify relevant variables that may interact in complex multivariate models to explain phenotypic variation. While several tools have been developed for assessing differential expression in sequence-based transcriptomics, the detection of statistical interactions between transcripts has received less attention in the area of RNA-seq analysis. We describe a new extension and assessment of Relief-F for feature selection in RNA-seq data. The ReliefSeq implementation adapts the number of nearest neighbors (k) for each gene to optimize the Relief-F test statistics (importance scores) for finding both main effects and interactions. We compare this gene-wise adaptive-k (gwak) Relief-F method with standard RNA-seq feature selection tools, such as DESeq and edgeR, and with the popular machine learning method Random Forests. We demonstrate performance on a panel of simulated data that have a range of distributional properties reflected in real mRNA-seq data including multiple transcripts with varying sizes of main effects and interaction effects. For simulated main effects, gwak-Relief-F feature selection performs comparably to standard tools DESeq and edgeR for ranking relevant transcripts. For gene-gene interactions, gwak-Relief-F outperforms all comparison methods at ranking relevant genes in all but the highest fold change/highest signal situations where it performs similarly. The gwak-Relief-F algorithm outperforms Random Forests for detecting relevant genes in all simulation experiments. In addition, Relief-F is comparable to the other methods based on computational time. We also apply ReliefSeq to an RNA-Seq study of smallpox vaccine to identify gene expression changes between vaccinia virus-stimulated and unstimulated samples. ReliefSeq is an attractive tool for inclusion in the suite of tools used for analysis of mRNA-Seq data; it has power to detect both main effects and interaction effects. Software Availability: http://insilico.utulsa.edu/ReliefSeq.php. PMID:24339943

Predicting Failure Under Laboratory Conditions: Learning the Physics of Slow Frictional Slip and Dynamic Failure

NASA Astrophysics Data System (ADS)

Rouet-Leduc, B.; Hulbert, C.; Riviere, J.; Lubbers, N.; Barros, K.; Marone, C.; Johnson, P. A.

2016-12-01

Forecasting failure is a primary goal in diverse domains that include earthquake physics, materials science, nondestructive evaluation of materials and other engineering applications. Due to the highly complex physics of material failure and limitations on gathering data in the failure nucleation zone, this goal has often appeared out of reach; however, recent advances in instrumentation sensitivity, instrument density and data analysis show promise toward forecasting failure times. Here, we show that we can predict frictional failure times of both slow and fast stick slip failure events in the laboratory. This advance is made possible by applying a machine learning approach known as Random Forests1(RF) to the continuous acoustic emission (AE) time series recorded by detectors located on the fault blocks. The RF is trained using a large number of statistical features derived from the AE time series signal. The model is then applied to data not previously analyzed. Remarkably, we find that the RF method predicts upcoming failure time far in advance of a stick slip event, based only on a short time window of data. Further, the algorithm accurately predicts the time of the beginning and end of the next slip event. The predicted time improves as failure is approached, as other data features add to prediction. Our results show robust predictions of slow and dynamic failure based on acoustic emissions from the fault zone throughout the laboratory seismic cycle. The predictions are based on previously unidentified tremor-like acoustic signals that occur during stress build up and the onset of macroscopic frictional weakening. We suggest that the tremor-like signals carry information about fault zone processes and allow precise predictions of failure at any time in the slow slip or stick slip cycle2. If the laboratory experiments represent Earth frictional conditions, it could well be that signals are being missed that contain highly useful predictive information. 1Breiman, L. Random forests. Machine Learning 45, 5-32 (2001). 2Rouet-Leduc, B. C. Hulbert, N. Lubbers, K. Barros and P. A. Johnson, Learning the physics of failure, in review (2016).

A Random Forest-based ensemble method for activity recognition.

PubMed

Feng, Zengtao; Mo, Lingfei; Li, Meng

2015-01-01

This paper presents a multi-sensor ensemble approach to human physical activity (PA) recognition, using random forest. We designed an ensemble learning algorithm, which integrates several independent Random Forest classifiers based on different sensor feature sets to build a more stable, more accurate and faster classifier for human activity recognition. To evaluate the algorithm, PA data collected from the PAMAP (Physical Activity Monitoring for Aging People), which is a standard, publicly available database, was utilized to train and test. The experimental results show that the algorithm is able to correctly recognize 19 PA types with an accuracy of 93.44%, while the training is faster than others. The ensemble classifier system based on the RF (Random Forest) algorithm can achieve high recognition accuracy and fast calculation.

Thrown object hazards in forest operations

Treesearch

Robert Rummer; John Klepac

2011-01-01

Mechanized equipment for forest operations provide better operator protection in this hazardous work environment. However operators of forestry cutting machines are now exposed to new hazards from the high-energy cutting devices used to cut trees and process logs. Anecdotal reports of thrown objects document a risk of injury and fatality. Two new ISO standards have...