Separate Support Vector Machine Research Articles

0537 Incident Hypertension Prediction in Obstructive Sleep Apnea using Machine Learning

Abstract Introduction Obstructive sleep apnea (OSA) is associated with hypertension due to intermittent hypoxia and sleep fragmentation. Due to the complex pathogenesis of hypertension, it is difficult to predict incident hypertension associated with OSA. A Machine Learning (ML) model to predict incident hypertension identified up to five years after the diagnosis of OSA by polysomnography developed. Methods Polysomnography provides time-series data on multiple physiological signals. We used the sleep heart health study (SHHS) cohort, where 4,797 participants had OSA. After excluding participants with pre-existing hypertension at baseline, the sample size was 2,652. 1,814 participants with follow-up data at 5 years were included (911/1,814, 50% with incident hypertension). In addition to clinical data (i.e. age and race), features extracted from polysomnography (heart rate variability, HRV calculated based on the electrocardiography R-R interval), electroencephalography delta power, statistical information (i.e., mean and standard deviation of signals), and heart rate periodicity functions fed to support vector machine (SVM) ML model to train and validate. The polysomnography features were calculated over the 30-second epochs identified based on respiratory events and EEG arousal and respiratory events annotation, and their corresponding parts in other signals based on sampling frequency. Technical artifacts in oxygen saturation and ECG were reconstructed with the interpolation method and removed from the signal respectively. The SVM is a robust ML method trained in an iterative fashion to find the global optimum. In comparison to the Deep Neural Network (DNN) approaches, SVMs results are interpretable. Each polysomnography signal and its corresponding features were trained on a separate SVM, followed by a fusion of the SVM results. The final results were fused by voting of individual SVM results. Results The SVM ML model thus far has achieved a test accuracy (area under the curve, AUC) of 66.06%, sensitivity 63.21%, and specificity 68.9%. Conclusion This proof-of-concept study suggests that supervised ML models, such as the SVM, may be useful in predicting incident hypertension associated with OSA. Further research is required regarding optimal input features to boost the accuracy, followed by external validation of the model in additional OSA cohorts. Support (if any) Research support 1R56HL157182, NIH/NHLBI

Automatic Detection of Genetic Diseases in Pediatric Age Using Pupillometry

Children with inherited retinal disorders have substantial vision impairment. They are divided into outer and inner retina illnesses, and they frequently result in infant blindness. Given the vast range of clinical and genetic symptoms associated with this type of sickness, the diagnosis might be difficult. causes (with more than 200 causal genes). It is frequently based on a complicated network of clinical tests, including invasive ones that are not necessarily suitable for newborns or young children. Thus, a new strategy is required that makes use of chromatic pupillometry, a method that is increasingly being employed to evaluate the inner and outer retinal functions. To assist in the detection of inherited retinal illnesses in pediatric children, this research offers a unique Clinical Decision Support System (CDSS) based on Machine Learning and employing Chromatic Pupillometry. The employment of specialized medical equipment (pupillometer) in conjunction with a specially created custom machine learning decision support system is suggested as a method that blends hardware and software. The features retrieved from the pupillometric data are classified using two separate Support Vector Machines (SVMs), one for each eye. Retinitis pigmentosa in children has been diagnosed using the specified CDSS. The system performed satisfactorily, as evidenced by the results of integrating the two SVMs into an ensemble model, which showed 91.176 accuracy, 1.0 sensitivity, and 0.0 specificity. This study is the first to use pupillometric data to apply machine learning.

Automatic Building Detection for Multi-Aspect SAR Images Based on the Variation Features

Multi-aspect synthetic aperture radar (SAR) images contain more information available for automatic target recognition (ATR) than images from a single view. However, the sensitivity to aspect angles also makes it hard to extract and integrate information from multi-aspect images. In this paper, we propose a novel method based on the variations features to realize automatic building detection in the image level. First, to get a comprehensive description of target variation patterns, statistical characteristic variances are derived from three representative and complementary categories. Then, these obtained features are fused and put in the K-means classifier for prescreening, whose results are used as the training sets in supervised classification later to avoid manual labeling. Second, for more precise detection performance, finer features in vector forms are obtained by principal component analysis (PCA). The variation patterns of these feature vectors are explored in two different manners of correlation and fluctuation analyses and processed by separate support vector machines (SVMs) after fusion. Finally, the independent SVM detection results are fused according to a maximum probability rule. Experiments conducted on two different airborne data sets demonstrate the robustness and effectiveness of the proposed method, in spite of significant target signature variabilities and cluttered background.

EnsembleFam: towards more accurate protein family prediction in the twilight zone

BackgroundCurrent protein family modeling methods like profile Hidden Markov Model (pHMM), k-mer based methods, and deep learning-based methods do not provide very accurate protein function prediction for proteins in the twilight zone, due to low sequence similarity to reference proteins with known functions.ResultsWe present a novel method EnsembleFam, aiming at better function prediction for proteins in the twilight zone. EnsembleFam extracts the core characteristics of a protein family using similarity and dissimilarity features calculated from sequence homology relations. EnsembleFam trains three separate Support Vector Machine (SVM) classifiers for each family using these features, and an ensemble prediction is made to classify novel proteins into these families. Extensive experiments are conducted using the Clusters of Orthologous Groups (COG) dataset and G Protein-Coupled Receptor (GPCR) dataset. EnsembleFam not only outperforms state-of-the-art methods on the overall dataset but also provides a much more accurate prediction for twilight zone proteins.ConclusionsEnsembleFam, a machine learning method to model protein families, can be used to better identify members with very low sequence homology. Using EnsembleFam protein functions can be predicted using just sequence information with better accuracy than state-of-the-art methods.

Robust ensemble of handcrafted and learned approaches for DNA-binding proteins

Purpose Automatic DNA-binding protein (DNA-BP) classification is now an essential proteomic technology. Unfortunately, many systems reported in the literature are tested on only one or two datasets/tasks. The purpose of this study is to create the most optimal and universal system for DNA-BP classification, one that performs competitively across several DNA-BP classification tasks. Design/methodology/approach Efficient DNA-BP classifier systems require the discovery of powerful protein representations and feature extraction methods. Experiments were performed that combined and compared descriptors extracted from state-of-the-art matrix/image protein representations. These descriptors were trained on separate support vector machines (SVMs) and evaluated. Convolutional neural networks with different parameter settings were fine-tuned on two matrix representations of proteins. Decisions were fused with the SVMs using the weighted sum rule and evaluated to experimentally derive the most powerful general-purpose DNA-BP classifier system. Findings The best ensemble proposed here produced comparable, if not superior, classification results on a broad and fair comparison with the literature across four different datasets representing a variety of DNA-BP classification tasks, thereby demonstrating both the power and generalizability of the proposed system. Originality/value Most DNA-BP methods proposed in the literature are only validated on one (rarely two) datasets/tasks. In this work, the authors report the performance of our general-purpose DNA-BP system on four datasets representing different DNA-BP classification tasks. The excellent results of the proposed best classifier system demonstrate the power of the proposed approach. These results can now be used for baseline comparisons by other researchers in the field.

Set of approaches based on 3D structure and position specific-scoring matrix for predicting DNA-binding proteins.

Because DNA-binding proteins (DNA-BPs) play a vital role in all aspects of genetic activity, the development of reliable and efficient systems for automatic DNA-BP classification is becoming a crucial proteomic technology. Key to this technology is the discovery of powerful protein representations and feature extraction methods. The goal of this article is to develop experimentally a system for automatic DNA-BP classification by comparing and combining different descriptors taken from different types of protein representations. The descriptors we evaluate include those starting from the position-specific scoring matrix (PSSM) of proteins, those derived from the amino-acid sequence (AAS), various matrix representations of proteins and features taken from the three-dimensional tertiary structure of proteins. We also introduce some new variants of protein descriptors. Each descriptor is used to train a separate support vector machine (SVM), and results are combined by sum rule. Our final system obtains state-or-the-art results on three benchmark DNA-BP datasets. The MATLAB code for replicating the experiments presented in this paper is available at https://github.com/LorisNanni.

Corrigendum to "On the integrity of functional brain networks in schizophrenia, Parkinson's disease, and advanced age: Evidence from connectivity-based single-subject classification."

Previous whole-brain functional connectivity studies achieved successful classifications of patients and healthy controls but only offered limited specificity as to affected brain systems. Here, we examined whether the connectivity patterns of functional systems affected in schizophrenia (SCZ), Parkinson's disease (PD), or normal aging equally translate into high classification accuracies for these conditions. We compared classification performance between pre-defined networks for each group and, for any given network, between groups. Separate support vector machine classifications of 86 SCZ patients, 80 PD patients, and 95 older adults relative to their matched healthy/young controls, respectively, were performed on functional connectivity in 12 task-based, meta-analytically defined networks using 25 replications of a nested 10-fold cross-validation scheme. Classification performance of the various networks clearly differed between conditions, as those networks that best classified one disease were usually non-informative for the other. For SCZ, but not PD, emotion-processing, empathy, and cognitive action control networks distinguished patients most accurately from controls. For PD, but not SCZ, networks subserving autobiographical or semantic memory, motor execution, and theory-of-mind cognition yielded the best classifications. In contrast, young-old classification was excellent based on all networks and outperformed both clinical classifications. Our pattern-classification approach captured associations between clinical and developmental conditions and functional network integrity with a higher level of specificity than did previous whole-brain analyses. Taken together, our results support resting-state connectivity as a marker of functional dysregulation in specific networks known to be affected by SCZ and PD, while suggesting that aging affects network integrity in a more global way. Hum Brain Mapp 38:5845-5858, 2017. © 2017 Wiley Periodicals, Inc.

On the integrity of functional brain networks in schizophrenia, Parkinson's disease, and advanced age: Evidence from connectivity-based single-subject classification.

Previous whole-brain functional connectivity studies achieved successful classifications of patients and healthy controls but only offered limited specificity as to affected brain systems. Here, we examined whether the connectivity patterns of functional systems affected in schizophrenia (SCZ), Parkinson's disease (PD), or normal aging equally translate into high classification accuracies for these conditions. We compared classification performance between pre-defined networks for each group and, for any given network, between groups. Separate support vector machine classifications of 86 SCZ patients, 80 PD patients, and 95 older adults relative to their matched healthy/young controls, respectively, were performed on functional connectivity in 12 task-based, meta-analytically defined networks using 25 replications of a nested 10-fold cross-validation scheme. Classification performance of the various networks clearly differed between conditions, as those networks that best classified one disease were usually non-informative for the other. For SCZ, but not PD, emotion-processing, empathy, and cognitive action control networks distinguished patients most accurately from controls. For PD, but not SCZ, networks subserving autobiographical or semantic memory, motor execution, and theory-of-mind cognition yielded the best classifications. In contrast, young-old classification was excellent based on all networks and outperformed both clinical classifications. Our pattern-classification approach captured associations between clinical and developmental conditions and functional network integrity with a higher level of specificity than did previous whole-brain analyses. Taken together, our results support resting-state connectivity as a marker of functional dysregulation in specific networks known to be affected by SCZ and PD, while suggesting that aging affects network integrity in a more global way. Hum Brain Mapp 38:5845-5858, 2017. © 2017 Wiley Periodicals, Inc.

Dynamic brain connectivity is a better predictor of PTSD than static connectivity.

Using resting-state functional magnetic resonance imaging, we test the hypothesis that subjects with post-traumatic stress disorder (PTSD) are characterized by reduced temporal variability of brain connectivity compared to matched healthy controls. Specifically, we test whether PTSD is characterized by elevated static connectivity, coupled with decreased temporal variability of those connections, with the latter providing greater sensitivity toward the pathology than the former. Static functional connectivity (FC; nondirectional zero-lag correlation) and static effective connectivity (EC; directional time-lagged relationships) were obtained over the entire brain using conventional models. Dynamic FC and dynamic EC were estimated by letting the conventional models to vary as a function of time. Statistical separation and discriminability of these metrics between the groups and their ability to accurately predict the diagnostic label of a novel subject were ascertained using separate support vector machine classifiers. Our findings support our hypothesis that PTSD subjects have stronger static connectivity, but reduced temporal variability of connectivity. Further, machine learning classification accuracy obtained with dynamic FC and dynamic EC was significantly higher than that obtained with static FC and static EC, respectively. Furthermore, results also indicate that the ease with which brain regions engage or disengage with other regions may be more sensitive to underlying pathology than the strength with which they are engaged. Future studies must examine whether this is true only in the case of PTSD or is a general organizing principle in the human brain. Hum Brain Mapp 38:4479-4496, 2017. © 2017 Wiley Periodicals, Inc.

Using measurements from wearable sensors for automatic scoring of Parkinson's disease motor states: Results from 7 patients.

The objective of this study was to investigate the validity of an objective gait measure for assessment of different motor states of advanced Parkinson's disease (PD) patients. Seven PD patients performed a gait task up to 15 times while wearing sensors on their upper and lower limbs. Each task was performed at specific points during a test day, following a single dose of levodopa-carbidopa. At the time of the tasks the patients were video recorded and three movement disorder experts rated their motor function on three clinical scales: a treatment response scale (TRS) that ranged from -3 (very bradykinetic) to 0 (ON) to +3 (very dyskinetic), a dyskinesia score that ranged from 0 (no dyskinesia) to 4 (extreme dyskinesia), and a bradykinesia score that ranged from 0 (no bradykinesia) to 4 (extreme bradykinesia). Raw accelerometer and gyroscope data of the sensors were processed and analyzed with time series analysis methods to extract features. The utilized features quantified separate limb movements as well as movement symmetries between the limbs. The features were processed with principal component analysis and the components were used as predictors for separate support vector machine (SVM) models for each of the three scales. The performance of each model was evaluated in a leave-one-patient out setting where the observations of a single patient were used as the testing set and the observations of the other 6 patients as the training set. Root mean square error (RMSE) and correlation coefficients for the predictions showed a good ability of the models to map the sensor data into the rating scales. There were strong correlations between the SVM models and the mean ratings of TRS (0.79; RMSE=0.70), bradykinesia score (0.79; RMSE=0.47), and bradykinesia score (0.78; RMSE=0.46). The results presented in this paper indicate that the use of wearable sensors when performing gait tasks can generate measurements that have a good correlation to subjective expert assessments.

Predicting early psychiatric readmission with natural language processing of narrative discharge summaries.

The ability to predict psychiatric readmission would facilitate the development of interventions to reduce this risk, a major driver of psychiatric health-care costs. The symptoms or characteristics of illness course necessary to develop reliable predictors are not available in coded billing data, but may be present in narrative electronic health record (EHR) discharge summaries. We identified a cohort of individuals admitted to a psychiatric inpatient unit between 1994 and 2012 with a principal diagnosis of major depressive disorder, and extracted inpatient psychiatric discharge narrative notes. Using these data, we trained a 75-topic Latent Dirichlet Allocation (LDA) model, a form of natural language processing, which identifies groups of words associated with topics discussed in a document collection. The cohort was randomly split to derive a training (70%) and testing (30%) data set, and we trained separate support vector machine models for baseline clinical features alone, baseline features plus common individual words and the above plus topics identified from the 75-topic LDA model. Of 4687 patients with inpatient discharge summaries, 470 were readmitted within 30 days. The 75-topic LDA model included topics linked to psychiatric symptoms (suicide, severe depression, anxiety, trauma, eating/weight and panic) and major depressive disorder comorbidities (infection, postpartum, brain tumor, diarrhea and pulmonary disease). By including LDA topics, prediction of readmission, as measured by area under receiver-operating characteristic curves in the testing data set, was improved from baseline (area under the curve 0.618) to baseline+1000 words (0.682) to baseline+75 topics (0.784). Inclusion of topics derived from narrative notes allows more accurate discrimination of individuals at high risk for psychiatric readmission in this cohort. Topic modeling and related approaches offer the potential to improve prediction using EHRs, if generalizability can be established in other clinical cohorts.

Pharmacologic attenuation of cross-modal sensory augmentation within the chronic pain insula.

Pain can be elicited through all mammalian sensory pathways yet cross-modal sensory integration, and its relationship to clinical pain, is largely unexplored. Centralized chronic pain conditions such as fibromyalgia are often associated with symptoms of multisensory hypersensitivity. In this study, female patients with fibromyalgia demonstrated cross-modal hypersensitivity to visual and pressure stimuli compared with age- and sex-matched healthy controls. Functional magnetic resonance imaging revealed that insular activity evoked by an aversive level of visual stimulation was associated with the intensity of fibromyalgia pain. Moreover, attenuation of this insular activity by the analgesic pregabalin was accompanied by concomitant reductions in clinical pain. A multivariate classification method using support vector machines (SVM) applied to visual-evoked brain activity distinguished patients with fibromyalgia from healthy controls with 82% accuracy. A separate SVM classification of treatment effects on visual-evoked activity reliably identified when patients were administered pregabalin as compared with placebo. Both SVM analyses identified significant weights within the insular cortex during aversive visual stimulation. These data suggest that abnormal integration of multisensory and pain pathways within the insula may represent a pathophysiological mechanism in some chronic pain conditions and that insular response to aversive visual stimulation may have utility as a marker for analgesic drug development.

A Robust Multimodal Biometric System Integrating Iris, Face and Fingerprint using Multiple SVMs

This paper presents a robust multimodal biometric recognition system integrating iris, face and fingerprint based on match score level fusion using multiple support vector machines (SVMs). Here, multiple support vector machines are applied in parallel fashion to overcome the problem of missing biometric traits. It considers every possible combination of all the three biometric traits (iris, face and fingerprint) individually. Each possible combination of biometric traits has a separate SVM to combine the available match scores to generate the final decision. Existing multimodal biometric recognition systems are based on the assumption that the set of biometric traits to be integrated is always present as a whole at the time of authentication. But sometimes it is not possible due to some unavoidable circumstances (e.g. injury may be caused, person may be under some medical treatment, corresponding trait may be missing etc.). The performance of the proposed system is evaluated on a public dataset demonstrating its recognition accuracy regarding FAR (False Accept Rate) and FRR (False Reject Rate). Keywords: support vector machine (SVM), score level fusion, iris recognition, face recognition, fingerprint recognition, receiver operating characteristic (ROC) curve.

An empirical study of different approaches for protein classification.

Many domains would benefit from reliable and efficient systems for automatic protein classification. An area of particular interest in recent studies on automatic protein classification is the exploration of new methods for extracting features from a protein that work well for specific problems. These methods, however, are not generalizable and have proven useful in only a few domains. Our goal is to evaluate several feature extraction approaches for representing proteins by testing them across multiple datasets. Different types of protein representations are evaluated: those starting from the position specific scoring matrix of the proteins (PSSM), those derived from the amino-acid sequence, two matrix representations, and features taken from the 3D tertiary structure of the protein. We also test new variants of proteins descriptors. We develop our system experimentally by comparing and combining different descriptors taken from the protein representations. Each descriptor is used to train a separate support vector machine (SVM), and the results are combined by sum rule. Some stand-alone descriptors work well on some datasets but not on others. Through fusion, the different descriptors provide a performance that works well across all tested datasets, in some cases performing better than the state-of-the-art.

Identification and characterization of plastid-type proteins from sequence-attributed features using machine learning

BackgroundPlastids are an important component of plant cells, being the site of manufacture and storage of chemical compounds used by the cell, and contain pigments such as those used in photosynthesis, starch synthesis/storage, cell color etc. They are essential organelles of the plant cell, also present in algae. Recent advances in genomic technology and sequencing efforts is generating a huge amount of DNA sequence data every day. The predicted proteome of these genomes needs annotation at a faster pace. In view of this, one such annotation need is to develop an automated system that can distinguish between plastid and non-plastid proteins accurately, and further classify plastid-types based on their functionality. We compared the amino acid compositions of plastid proteins with those of non-plastid ones and found significant differences, which were used as a basis to develop various feature-based prediction models using similarity-search and machine learning.ResultsIn this study, we developed separate Support Vector Machine (SVM) trained classifiers for characterizing the plastids in two steps: first distinguishing the plastid vs. non-plastid proteins, and then classifying the identified plastids into their various types based on their function (chloroplast, chromoplast, etioplast, and amyloplast). Five diverse protein features: amino acid composition, dipeptide composition, the pseudo amino acid composition, Nterminal-Center-Cterminal composition and the protein physicochemical properties are used to develop SVM models. Overall, the dipeptide composition-based module shows the best performance with an accuracy of 86.80% and Matthews Correlation Coefficient (MCC) of 0.74 in phase-I and 78.60% with a MCC of 0.44 in phase-II. On independent test data, this model also performs better with an overall accuracy of 76.58% and 74.97% in phase-I and phase-II, respectively. The similarity-based PSI-BLAST module shows very low performance with about 50% prediction accuracy for distinguishing plastid vs. non-plastids and only 20% in classifying various plastid-types, indicating the need and importance of machine learning algorithms.ConclusionThe current work is a first attempt to develop a methodology for classifying various plastid-type proteins. The prediction modules have also been made available as a web tool, PLpred available at http://bioinfo.okstate.edu/PLpred/ for real time identification/characterization. We believe this tool will be very useful in the functional annotation of various genomes.

A fast and accurate method for large-scale transmembrane beta barrel topology prediction

Transmembrane beta barrel proteins (TMBs) play a major role in the normal functioning of the cell and are an important constituent of the translocation machinery of the outer membrane proteins in bacteria, mitochondria and chloroplast. Currently there are only around 36 experimental 3D structures available for TMBs in PDB (at 30% sequence identity). Furthermore, with large amounts of sequence data available from high-throughput methods, it is imperative to develop accurate and fast computational methods for their identification and topology prediction. Here, we present a concept for a fast method to identify TMBs and predict their topology. The method uses sparse encoded amino acid data as input and employs a Support Vector Machine (SVM) and a Hidden Markov Model (HMM) to generate accurate topologies. The topologies in the training phase are divided into pre-barrel state, outer-loop state, inner-loop state and the transmembrane beta-strand state. In the first stage, 4 separate SVMs are used to predict the local state preference for each residue. A profile generated from the probabilities thus obtained is used as input to the HMM stage to determine the overall topology. If the number of predicted strands is between 8 and 24, then the given sequence is identified as a TMB. We see the application of our method in the proteome-wide topology prediction of TMBs, where current methods might have a limitation due to the time consuming homologous sequence search step. Funding Christoph Peters: Travel Fellowship awarded by ISCB Student Council with financial support from EMBO, 2012 J-A Ekstroms travel stipend, 2012 A fast and accurate method for large-scale transmembrane beta barrel topology prediction

Automatic discrimination between safe and unsafe swallowing using a reputation-based classifier

BackgroundSwallowing accelerometry has been suggested as a potential non-invasive tool for bedside dysphagia screening. Various vibratory signal features and complementary measurement modalities have been put forth in the literature for the potential discrimination between safe and unsafe swallowing. To date, automatic classification of swallowing accelerometry has exclusively involved a single-axis of vibration although a second axis is known to contain additional information about the nature of the swallow. Furthermore, the only published attempt at automatic classification in adult patients has been based on a small sample of swallowing vibrations.MethodsIn this paper, a large corpus of dual-axis accelerometric signals were collected from 30 older adults (aged 65.47 ± 13.4 years, 15 male) referred to videofluoroscopic examination on the suspicion of dysphagia. We invoked a reputation-based classifier combination to automatically categorize the dual-axis accelerometric signals into safe and unsafe swallows, as labeled via videofluoroscopic review. From these participants, a total of 224 swallowing samples were obtained, 164 of which were labeled as unsafe swallows (swallows where the bolus entered the airway) and 60 as safe swallows. Three separate support vector machine (SVM) classifiers and eight different features were selected for classification.ResultsWith selected time, frequency and information theoretic features, the reputation-based algorithm distinguished between safe and unsafe swallowing with promising accuracy (80.48 ± 5.0%), high sensitivity (97.1 ± 2%) and modest specificity (64 ± 8.8%). Interpretation of the most discriminatory features revealed that in general, unsafe swallows had lower mean vibration amplitude and faster autocorrelation decay, suggestive of decreased hyoid excursion and compromised coordination, respectively. Further, owing to its performance-based weighting of component classifiers, the static reputation-based algorithm outperformed the democratic majority voting algorithm on this clinical data set.ConclusionGiven its computational efficiency and high sensitivity, reputation-based classification of dual-axis accelerometry ought to be considered in future developments of a point-of-care swallow assessment where clinical informatics are desired.

Automatic extraction of metadata from scientific publications for CRIS systems

PurposeThe aim of this paper is to develop a system for automatic extraction of metadata from scientific papers in PDF format for the information system for monitoring the scientific research activity of the University of Novi Sad (CRIS UNS).Design/methodology/approachThe system is based on machine learning and performs automatic extraction and classification of metadata in eight pre‐defined categories. The extraction task is realised as a classification process. For the purpose of classification each row of text is represented with a vector that comprises different features: formatting, position, characteristics related to the words, etc. Experiments were performed with standard classification models. Both a single classifier with all eight categories and eight individual classifiers were tested. Classifiers were evaluated using the five‐fold cross validation, on a manually annotated corpus comprising 100 scientific papers in PDF format, collected from various conferences, journals and authors' personal web pages.FindingsBased on the performances obtained on classification experiments, eight separate support vector machines (SVM) models (each of which recognises its corresponding category) were chosen. All eight models were established to have a good performance. The F‐measure was over 85 per cent for almost all of the classifiers and over 90 per cent for most of them.Research limitations/implicationsAutomatically extracted metadata cannot be directly entered into CRIS UNS but requires control of the curators.Practical implicationsThe proposed system for automatic metadata extraction using support vector machines model was integrated into the software system, CRIS UNS. Metadata extraction has been tested on the publications of researchers from the Department of Mathematics and Informatics of the Faculty of Sciences in Novi Sad. Analysis of extracted metadata from these publications showed that the performance of the system for the previously unseen data is in accordance with that obtained by the cross‐validation from eight separate SVM classifiers. This system will help in the process of synchronising metadata from CRIS UNS with other institutional repositories.Originality/valueThe paper documents a fully automated system for metadata extraction from scientific papers that was developed. The system is based on the SVM classifier and open source tools, and is capable of extracting eight types of metadata from scientific articles of any format that can be converted to PDF. Although developed as part of CRIS UNS, the proposed system can be integrated into other CRIS systems, as well as institutional repositories and library management systems.

Using Procedural Codes to Supplement Risk Adjustment: A Nonparametric Learning Approach

The American College of Surgeons National Surgical Quality Improvement Program collects information related to procedures in the form of the work relative value unit (RVU) and current procedural terminology (CPT) code. We propose and evaluate a fully automated nonparametric learning approach that maps individual CPT codes to perioperative risk. National Surgical Quality Improvement Program participant use file data for 2005-2006 were used to develop 2 separate support vector machines (SVMs) to learn the relationship between CPT codes and 30-day mortality or morbidity. SVM parameters were determined using cross-validation. SVMs were evaluated on participant use file data for 2007 and 2008. Areas under the receiver operating characteristic curve (AUROCs) were each compared with the respective AUROCs for work RVU and for standard CPT categories. We then compared the AUROCs for multivariable models, including preoperative variables, RVU, and CPT categories, with and without the SVM operation scores. SVM operation scores had AUROCs between 0.798 and 0.822 for mortality and between 0.745 and 0.758 for morbidity on the participant use file used for both training (2005-2006) and testing (2007 and 2008). This was consistently higher than the AUROCs for both RVU and standard CPT categories (p < 0.001). AUROCs of multivariable models were higher for 30-day mortality and morbidity when SVM operation scores were included. This difference was not significant for mortality but statistically significant, although small, for morbidity. Nonparametric methods from artificial intelligence can translate CPT codes to aid in the assessment of perioperative risk. This approach is fully automated and can complement the use of work RVU or traditional CPT categories in multivariable risk adjustment models like the National Surgical Quality Improvement Program.

Estimating soil moisture using remote sensing data: A machine learning approach

Soil moisture is an integral quantity in hydrology that represents the average conditions in a finite volume of soil. In this paper, a novel regression technique called Support Vector Machine (SVM) is presented and applied to soil moisture estimation using remote sensing data. SVM is based on statistical learning theory that uses a hypothesis space of linear functions based on Kernel approach. SVM has been used to predict a quantity forward in time based on training from past data. The strength of SVM lies in minimizing the empirical classification error and maximizing the geometric margin by solving inverse problem. SVM model is applied to 10 sites for soil moisture estimation in the Lower Colorado River Basin (LCRB) in the western United States. The sites comprise low to dense vegetation. Remote sensing data that includes backscatter and incidence angle from Tropical Rainfall Measuring Mission (TRMM), and Normalized Difference Vegetation Index (NDVI) from Advanced Very High Resolution Radiometer (AVHRR) are used to estimate soil water content (SM). Simulated SM (%) time series for the study sites are available from the Variable Infiltration Capacity Three Layer (VIC) model for top 10 cm layer of soil for the years 1998–2005. SVM model is trained on 5 years of data, i.e. 1998–2002 and tested on 3 years of data, i.e. 2003–2005. Two models are developed to evaluate the strength of SVM modeling in estimating soil moisture. In model I, training and testing are done on six sites, this results in six separate SVM models – one for each site. Model II comprises of two subparts: (a) data from all six sites used in model I is combined and a single SVM model is developed and tested on same sites and (b) a single model is developed using data from six sites (same as model II-A) but this model is tested on four separate sites not used to train the model. Model I shows satisfactory results, and the SM estimates are in good agreement with the estimates from VIC model. The SM estimate correlation coefficients range from 0.34 to 0.77 with RMSE less than 2% at all the selected sites. A probabilistic absolute error between the VIC SM and modeled SM is computed for all models. For model I, the results indicate that 80% of the SM estimates have an absolute error of less than 5%, whereas for model II-A and II-B, 80% and 60% of the SM estimates have an error less than 10% and 15%, respectively. SVM model is also trained and tested for measured soil moisture in the LCRB. Results with RMSE, MAE and R of 2.01, 1.97, and 0.57, respectively show that the SVM model is able to capture the variability in measured soil moisture. Results from the SVM modeling are compared with the estimates obtained from feed forward-back propagation Artificial Neural Network model (ANN) and Multivariate Linear Regression model (MLR); and show that SVM model performs better for soil moisture estimation than ANN and MLR models.