Types Of Classification Models Research Articles

On Measurement Validity and Language Models: Increasing Validity and Decreasing Bias with Instructions

ABSTRACT Language models like BERT or GPT are becoming increasingly popular measurement tools, but are the measurements they produce valid? Literature suggests that there is still a relevant gap between the ambitions of computational text analysis methods and the validity of their outputs. One prominent threat to validity is hidden biases in the training data, where models learn group-specific language patterns instead of the concept researchers want to measure. This paper investigates to what extent these biases impact the validity of measurements created with language models. We conduct a comparative analysis across nine group types in four datasets with three types of classification models, focusing on the robustness of models against biases and on the validity of their outputs. While we find that all types of models learn biases, the effects on validity are surprisingly small. In particular when models receive instructions as an additional input, they become more robust against biases from the fine-tuning data and produce more valid measurements across different groups. An instruction-based model (BERT-NLI) sees its average test-set performance decrease by only 0.4% F1 macro when trained on biased data and its error probability on groups it has not seen during training increases only by 0.8%.

A Comparative Analysis of Machine Learning Models for the Detection of Undiagnosed Diabetes Patients

Introduction: Early detection of type 2 diabetes is essential for preventing long-term complications. However, screening the entire population for diabetes is not cost-effective, so identifying individuals at high risk for this disease is crucial. The aim of this study was to compare the performance of five diverse machine learning (ML) models in classifying undiagnosed diabetes using large heterogeneous datasets. Methods: We used machine learning data from several years of the National Health and Nutrition Examination Survey (NHANES) from 2005 to 2018 to identify people with undiagnosed diabetes. The dataset included 45,431 participants, and biochemical confirmation of glucose control (HbA1c) were used to identify undiagnosed diabetes. The predictors were based on simple and clinically obtainable variables, which could be feasible for prescreening for diabetes. We included five ML models for comparison: random forest, AdaBoost, RUSBoost, LogitBoost, and a neural network. Results: The prevalence of undiagnosed diabetes was 4%. For the classification of undiagnosed diabetes, the area under the ROC curve (AUC) values were between 0.776 and 0.806. The positive predictive values (PPVs) were between 0.083 and 0.091, the negative predictive values (NPVs) were between 0.984 and 0.99, and the sensitivities were between 0.742 and 0.871. Conclusion: We have demonstrated that several types of classification models can accurately classify undiagnosed diabetes from simple and clinically obtainable variables. These results suggest that the use of machine learning for prescreening for undiagnosed diabetes could be a useful tool in clinical practice.

Utilizing MATLAB machine learning models to categorize transient events in a nuclear power plant using generic pressurized water reactor simulator

Enhancing safety and dependability within nuclear power facilities holds paramount importance in safeguarding both individuals and the environment. The adoption of machine learning for diagnosing faults in these plants is steadily gaining interest, driven by its capacity to detect faults, alleviate human errors in high-pressure scenarios, and ensure the secure and consistent operation of these facilities swiftly and accurately.This paper examines the use of machine learning models for fault diagnostics, specifically, the identification of transient events in a nuclear power plant to reduce human errors. The data was collected from WSC’s Generic Pressurized Water Reactor (GPWR) simulator and processed using MATLAB. The simulator encompasses models for both the primary and secondary systems of the nuclear power plant (NPP). Additionally, it incorporates models for the control systems and instrumentation responsible for monitoring and regulating the reactor, serving as integral components for data extraction and transient modeling. A total of 9 different transient events were simulated with 12 different initial conditions to create a dataset with 72,000 observations. Nine types of classification models (33 total preset models) were trained and validated using the classification learners application. Among them, the Neural Network Classifiers (NNC) displayed the highest average accuracy of 90%. The Fine Tree, Ensemble Bagged Trees, and Medium Neural Network models were the best-performing individual models with validation accuracies above 90% and a maximum training time of 8 min. These models were further analyzed using accuracy, confusion matrix, precision, recall, and F1 score. To optimize these models, techniques such as different validation schemes and feature selection were utilized to further reduce their training time and improve their prediction accuracy. The optimized models boasted comparable accuracies with a maximum training time of under 1.5 min. The results of this study exhibited favorable comparisons with other machine-learning endeavors in the field of reactor transient detection and diagnostics. Notably, the study maintained low execution and computation times while preserving high levels of accuracy. This study offers insightful information on how AI and machine learning can be used to improve nuclear power plant diagnostics, enhance safety, and provide support to the operator.

Application and interpretation of machine learning models in predicting the risk of severe obstructive sleep apnea in adults

BackgroundObstructive sleep apnea (OSA) is a globally prevalent disease with a complex diagnostic method. Severe OSA is associated with multi-system dysfunction. We aimed to develop an interpretable machine learning (ML) model for predicting the risk of severe OSA and analyzing the risk factors based on clinical characteristics and questionnaires.MethodsThis was a retrospective study comprising 1656 subjects who presented and underwent polysomnography (PSG) between 2018 and 2021. A total of 23 variables were included, and after univariate analysis, 15 variables were selected for further preprocessing. Six types of classification models were used to evaluate the ability to predict severe OSA, namely logistic regression (LR), gradient boosting machine (GBM), extreme gradient boosting (XGBoost), adaptive boosting (AdaBoost), bootstrapped aggregating (Bagging), and multilayer perceptron (MLP). All models used the area under the receiver operating characteristic curve (AUC) was calculated as the performance metric. We also drew SHapley Additive exPlanations (SHAP) plots to interpret predictive results and to analyze the relative importance of risk factors. An online calculator was developed to estimate the risk of severe OSA in individuals.ResultsAmong the enrolled subjects, 61.47% (1018/1656) were diagnosed with severe OSA. Multivariate LR analysis showed that 10 of 23 variables were independent risk factors for severe OSA. The GBM model showed the best performance (AUC = 0.857, accuracy = 0.766, sensitivity = 0.798, specificity = 0.734). An online calculator was developed to estimate the risk of severe OSA based on the GBM model. Finally, waist circumference, neck circumference, the Epworth Sleepiness Scale, age, and the Berlin questionnaire were revealed by the SHAP plot as the top five critical variables contributing to the diagnosis of severe OSA. Additionally, two typical cases were analyzed to interpret the contribution of each variable to the outcome prediction in a single patient.ConclusionsWe established six risk prediction models for severe OSA using ML algorithms. Among them, the GBM model performed best. The model facilitates individualized assessment and further clinical strategies for patients with suspected severe OSA. This will help to identify patients with severe OSA as early as possible and ensure their timely treatment.Trial registrationRetrospectively registered.

Optimal Classification of Emotions from Electroencephalography (EEG) Signals

It is incredibly challenging to build an intelligent algorithm for emotion recognition that can deliver high accuracy because electroencephalography (EEG) signals are not stationary, nonlinear, and noisy. First, decomposing the preprocessed EEG signals of the SEED dataset into five frequency bands: delta, theta, alpha, beta, and gamma, and then calculated their energy and entropy from the extracted features. Then Principal Component Analysis (PCA) method for feature reduction was performed. It is important to note that different types of wavelets transform (db6, db5, etc.) were tested, and hyperparameter tuning of classification models was done to obtain optimal accuracy. The next step is classifying the emotions into three states: -1(negative), 0(neutral), and +1(positive), and tested the dataset on two types of classification models, namely Random Forest and Support vector machine (SVM). SVM gives better performance compared to Random Forest with an accuracy of 80.74%.

Testing the potential of the Sow Stance Information System (SowSIS) based on a force plate system built into an electronic sow feeder for on-farm automatic lameness detection in breeding sows

Lameness is a common problem in breeding sows, which often goes undetected for long periods with severe consequences for animal welfare and farm productivity. Automatic lameness detection could help pig farmers to recognise and treat lameness sooner. The SowSIS (Sow Stance Information System) is a device consisting of four force plates and providing non-invasive force measurements per leg of the sow. In this study, the SowSIS was built into electronic sow feeders and validated for lameness detection in group-housed gestating sows. Data was automatically collected for 71 sows. Visual gait scoring was performed twice a week using a 150-mm tagged visual analogue scale to determine the sows' lameness status. Only data from 32 gait scoring days were included, adding up to 674 sow days. A sow was classified as lame using >60 mm as the cut-off value for the visual gait scores. Stance variables were calculated from the SowSIS data per sow per day. First, a multivariable linear mixed model was used to detect lameness, using stance variables with significant influence on the gait score. The model's performance was 78.5% sensitivity, 81.4% specificity, 80.7% accuracy, 57.4% lame predictive value and 92.2% non-lame predictive value. Second, five types of classification models were tested to determine the lame leg on a sub-dataset. The random forest model could predict the lame leg correctly 90% of the time. The SowSIS shows great promise as an on-farm lameness detection system, as it allows continuous non-invasive data collection in a practical setting. • The Sow Stance Information System is robust and well adapted for on-farm use. • Daily non-invasive stance measurements of individual sows during feeding visits. • A linear mixed model using stance variables can correctly detect lameness in sows. • The lame leg can be identified using machine learning models.

Cost-sensitive probability for weighted voting in an ensemble model for multi-class classification problems

Ensemble learning is an algorithm that utilizes various types of classification models. This algorithm can enhance the prediction efficiency of component models. However, the efficiency of combining models typically depends on the diversity and accuracy of the predicted results of ensemble models. However, the problem of multi-class data is still encountered. In the proposed approach, cost-sensitive learning was implemented to evaluate the prediction accuracy for each class, which was used to construct a cost-sensitivity matrix of the true positive (TP) rate. This TP rate can be used as a weight value and combined with a probability value to drive ensemble learning for a specified class. We proposed an ensemble model, which was a type of heterogenous model, namely, a combination of various individual classification models (support vector machine, Bayes, K-nearest neighbour, naïve Bayes, decision tree, and multi-layer perceptron) in experiments on 3-, 4-, 5- and 6-classifier models. The efficiencies of the propose models were compared to those of the individual classifier model and homogenous models (Adaboost, bagging, stacking, voting, random forest, and random subspaces) with various multi-class data sets. The experimental results demonstrate that the cost-sensitive probability for the weighted voting ensemble model that was derived from 3 models provided the most accurate results for the dataset in multi-class prediction. The objective of this study was to increase the efficiency of predicting classification results in multi-class classification tasks and to improve the classification results.

PBGAN: Learning PPG Representations From GAN for Time-Stable and Unique Verification System

Photoplethysmography (PPG) is a non-invasive physiological signal that captures the changes in blood volume resulted from heart activity. It carries unique person-specific characteristics that can be utilized for biometric systems. Currently, the use of a biometric system is paramount to ensure the security of the user’s identity. Due to the high sensitivity of the PPG signal, it suffers from extreme variations within the same subject when obtained at different time instances. These variations impose a challenge to employ the PPG signal and hinder the algorithm generalization for many applications including verification and identification systems. In this work, we propose a PPG Biometric Generative Adversarial Network (PBGAN) to create synthetic person-specific and time-stable PPG signals for genuine samples. Two types of classification models are employed with the PBGAN where the focus is on verification scenarios. In addition, we expand our previously recorded PPG dataset from 100 to 170 participants where the new size guarantees the generalization capability of the proposed system. This database and another three public ones are employed to evaluate the performances in terms of uniqueness and time stability. Furthermore, we consider three different training strategies to simulate practical scenarios. The best results acquired from our collected database in terms of Equal Error Rate (EER) is 1.3% for the single-session and 11.5% for the two-sessions scenarios which demonstrate the effectiveness of the proposed method in improving the verification system’s performance. Compared to our previous work, we achieve 1.3% and 1.4% EER improvements in two-sessions’ databases with small computational times which reveals the superiority of our proposed approach for real applications. Later, the code and dataset can be accessed in https://github.com/eoduself .

Stress Classification Using Photoplethysmogram-Based Spatial and Frequency Domain Images

Stress is subjective and is manifested differently from one person to another. Thus, the performance of generic classification models that classify stress status is crude. Building a person-specific model leads to a reliable classification, but it requires the collection of new data to train a new model for every individual and needs periodic upgrades because stress is dynamic. In this paper, a new binary classification (called stressed and non-stressed) approach is proposed for a subject’s stress state in which the inter-beat intervals extracted from a photoplethysomogram (PPG) were transferred to spatial images and then to frequency domain images according to the number of consecutive. Then, the convolution neural network (CNN) was used to train and validate the classification accuracy of the person’s stress state. Three types of classification models were built: person-specific models, generic classification models, and calibrated-generic classification models. The average classification accuracies achieved by person-specific models using spatial images and frequency domain images were 99.9%, 100%, and 99.8%, and 99.68%, 98.97%, and 96.4% for the training, validation, and test, respectively. By combining 20% of the samples collected from test subjects into the training data, the calibrated generic models’ accuracy was improved and outperformed the generic performance across both the spatial and frequency domain images. The average classification accuracy of 99.6%, 99.9%, and 88.1%, and 99.2%, 97.4%, and 87.6% were obtained for the training set, validation set, and test set, respectively, using the calibrated generic classification-based method for the series of inter-beat interval (IBI) spatial and frequency domain images. The main contribution of this study is the use of the frequency domain images that are generated from the spatial domain images of the IBI extracted from the PPG signal to classify the stress state of the individual by building person-specific models and calibrated generic models.

Technology choices in public transport planning: A classification framework

Choice of public transport technologies in cities is not straightforward: while the academy focuses on optimization models to determine which modes should a specific city have, policy makers rely on simple recommendations which are based on city population and income. We estimated six types of classification models that could allow for more precise recommendations yet are simple enough to be applied by the authorities. We considered typical variables as population and Gross Domestic Product of cities but also geographic and morphologic characteristics in a database of 400 cities from North and South America. Ordered Probit and Multinomial Logit models were the most accurate, with a success rate over 80% in the validation subset. Among the explanatory variables, city population and GDP per capita were as expected the most significant, but fare integration, car ownership and city shape were also relevant. Even if existent public transport modes in cities are not necessarily optimal, the classification models developed can give an insight for policy makers, in the sense that cities whose public transportation complexity cannot be explained by the models are more likely to have a suboptimal public transportation system.

Automated species identification of frog choruses in environmental recordings using acoustic indices

Acoustic monitoring provides opportunities for scaling up bioacoustic study of vocal animals to greater temporal and spatial scales. However, the large amounts of audio that can be easily and efficiently collected necessitates automated methods of analysis to extract useful ecological data. Acoustic indices have been used in spectrographic visualisation of long environmental recordings to successfully identify many biological sounds from their acoustic patterns and features. In particular, the choruses of several frog species are conspicuous in these spectrogram images which suggests that acoustic indices may be useful for detecting species in automated sound classification algorithms. The aim of this study was to investigate the use of acoustic indices as predictors in classification models for automated identification of frog species in environmental sound recordings from breeding habitats in north Queensland, Australia. Three types of classification models (random forests, support vector machines and gradient boosting) were trained and validated on a data set of 3274 1-minute audio segments labelled for the presence or absence of calling of 12 target frog species, and a feature set of 11 acoustic indices calculated on frequency bins of bandwidth 43.1 Hz. Classification performance was high for all 12 target species on the validation data set held out from the labelled training data (precision range 0.90–1.00 and recall range 0.83–0.99). However, performance declined for most target species when predicting frog calling on a further test data set taken from unseen recordings from the same sites. Best prediction results on the test data were achieved for species with the most training data, indicating accuracy may be improved by increasing training data, and this method is best suited to predicting chorusing of common species.

Transportation Mode Recognition With Deep Forest Based on GPS Data

Transportation mode recognition (TMR) is a common but critical task in the human behavior research field, which provides decision support for urban traffic planning, public facility arrangement, travel route recommendations, etc. The rapid development of urban information technology, mobile sensors and artificial intelligence has generated solutions for TMR; however, they rely on extra sensors and Geographic Information System (GIS) information, which are not always available. Recognition is usually simplified by disregarding the trajectories among transportation mode change points. In this paper, we proposed an ensemble learning-based approach to automatically recognize transportation modes (including a hybrid mode) using only Global Positioning System (GPS) data. A total of 72 features were extracted to better distinguish different transportation modes. Furthermore, we exploited a deep forest to combine various types of classification models, which facilitates robust learning with different trajectory samples and modes. The experimental results for the Geolife dataset show the efficiency of our approach, and the improved deep forest model achieved the best performance among all experiments that we conducted with 88.6% accuracy.

Application of machine learning to predict obstructive sleep apnea syndrome severity

Obstructive sleep apnea syndrome has become an important public health concern. Polysomnography is traditionally considered an established and effective diagnostic tool providing information on the severity of obstructive sleep apnea syndrome and the degree of sleep fragmentation. However, the numerous steps in the polysomnography test to diagnose obstructive sleep apnea syndrome are costly and time consuming. This study aimed to test the efficacy and clinical applicability of different machine learning methods based on demographic information and questionnaire data to predict obstructive sleep apnea syndrome severity. We collected data about demographic characteristics, spirometry values, gas exchange (PaO2, PaCO2) and symptoms (Epworth Sleepiness Scale, snoring, etc.) of 313 patients with previous diagnosis of obstructive sleep apnea syndrome. After principal component analysis, we selected 19 variables which were used for further preprocessing and to eventually train seven types of classification models and five types of regression models to evaluate the prediction ability of obstructive sleep apnea syndrome severity, represented either by class or by apnea-hypopnea index. All models are trained with an increasing number of features and the results are validated through stratified 10-fold cross validation. Comparative results show the superiority of support vector machine and random forest models for classification, while support vector machine and linear regression are better suited to predict apnea-hypopnea index. Also, a limited number of features are enough to achieve the maximum predictive accuracy. The best average classification accuracy on test sets is 44.7 percent, with the same average sensitivity (recall). In only 5.7 percent of cases, a severe obstructive sleep apnea syndrome (class 4) is misclassified as mild (class 2). Regression results show a minimum achieved root mean squared error of 22.17. The problem of predicting apnea-hypopnea index or severity classes for obstructive sleep apnea syndrome is very difficult when using only data collected prior to polysomnography test. The results achieved with the available data suggest the use of machine learning methods as tools for providing patients with a priority level for polysomnography test, but they still cannot be used for automated diagnosis.

Predicting bankruptcy of manufacturing companies in EU

Article focuses on the prediction of bankruptcy of the 1,000 medium-sized retail business companies in EU from which 170 companies gone bankrupt in 2014 with respect to lag of the used features. In recent times, bankruptcy of manufacturing companies rapidly increased due to the impact of the recession, which produces economic and social problems accordingly. Therefore, the need for bankruptcy prediction models is very high. From various types of classifi cation models we chose Support vector machines method with spline, hyperbolic tangent and RBF ANOVA kernels, Decision trees, Random forests and Adaptive boosting to acquire best results. Pre-processing is enhanced with fi lter based feature selection like Gain ratio and Relief algorithm to acquire attributes with the best information value. As we can see both fi ltering methods offers different variables to be used in the classifi cation and Decision trees wrapper algorithm chose less number than its competitors. Suitable attributes as ROA, Interest cover, Solvency ratio based on assets and Operating revenues were mostly used but it also changes across the time, which are probably very obtainable. It is apparent that inappropriate theoretical value of one variable does not necessarily lead to bankruptcy, so it is better to use combinations of these variables. From the results it is obvious that with the rising distance to the bankruptcy there drops precision of bankruptcy prediction. The last year (2013) with avaible fi nancial data offers best total prediction accuracy, thus we also infer both the Error I and II types for better recognizance of misclassifi cation rates. The Random forest and Decision trees offer better accuracy for bankruptcy prediction than SVM method, both method offers prediction accuracy which is comparable to previous empirical studies.

Dempster-Shafer theory for combining in silico evidence and estimating uncertainty in chemical risk assessment

In safety and risk assessment, various sources of evidence can be used to predict whether a given chemical compound may pose a risk for a particular type of toxicity. Evidence may come from computational approaches, such as quantitative structure-activity relationship (QSAR) models, rule-based structural alerts, or experimental data from assays relevant to the toxicity endpoint. Dempster-Shafer theory (DST) is a rigorous decision-theory approach that provides a way to generate predictions, estimate the uncertainty associated with each prediction, and combine multiple sources of evidence to obtain a weight-of-evidence prediction by quantitatively accounting for the reliability of each of the sources being combined. The general approach is presented for different types of classification models with illustrative examples for binary, ordinal, and multinomial classification. Application of the DST approach to model skin sensitization hazard potential based on local lymph node assay (LLNA) experimental data is presented.

Diagnosis of Missed Ductile Iron Melts with Process Modelling

AbstractThe paper presents an application of advanced data-driven (soft) models in finding the most probable particular causes of missed ductile iron melts. The proposed methodology was tested using real foundry data set containing 1020 records with contents of 9 chemical elements in the iron as the process input variables and the ductile iron grade as the output. This dependent variable was of discrete (nominal) type with four possible values: ‘400/18’, ‘500/07’, ‘500/07 special’ and ‘non-classified’, i.e. the missed melt. Several types of classification models were built and tested: MLP-type Artificial Neural Network, Support Vector Machine and two versions of Classification Trees. The best accuracy of predictions was achieved by one of the Classification Tree model, which was then used in the simulations leading to conversion of the missed melts to the expected grades. Two strategies of changing the input values (chemical composition) were tried: content of a single element at a time and simultaneous changes of a selected pair of elements. It was found that in the vast majority of the missed melts the changes of single elements concentrations have led to the change from the non-classified iron to its expected grade. In the case of the three remaining melts the simultaneous changes of pairs of the elements’ concentrations appeared to be successful and that those cases were in agreement with foundry staff expertise. It is concluded that utilizing an advanced data-driven process model can significantly facilitate diagnosis of defective products and out-of-control foundry processes.

Predicting financial distress of agriculture companies in EU

The objective of this paper is the prediction of financial distress (default of payment or insolvency) of 250 agriculture business companies in the EU from which 62 companies defaulted in 2014 with respect to lag of the used attributes. From many types of classification models, there was chosen the Logistic regression, the Support vector machines method with the RBF ANOVA kernel, the Decision Trees and the Adaptive Boosting based on the decision trees to acquire the best results. From the results, it is obvious that with the increasing distance to the bankruptcy, there decreases the average accuracy of the financial distress prediction and there is a greater difference between the active and distressed companies in terms of liquidity, rentability and debt ratios. The Decision trees and Adaptive Boosting offer a better accuracy for the distress prediction than the SVM and logit methods, what is comparable to the previous studies. From the total of 15 accounting variables, there were constructed classification trees by the Decision Trees with the inner feature selection method for the better visualization, what reduces the full data set only to 1 or 2 attributes: ROA and Long-term Debt to Total Assets Ratio in 2011, ROA and Current Ratio in 2012, ROA in 2013 for the discrimination of the distressed companies.

Improved supervised classification of accelerometry data to distinguish behaviors of soaring birds.

Soaring birds can balance the energetic costs of movement by switching between flapping, soaring and gliding flight. Accelerometers can allow quantification of flight behavior and thus a context to interpret these energetic costs. However, models to interpret accelerometry data are still being developed, rarely trained with supervised datasets, and difficult to apply. We collected accelerometry data at 140Hz from a trained golden eagle (Aquila chrysaetos) whose flight we recorded with video that we used to characterize behavior. We applied two forms of supervised classifications, random forest (RF) models and K-nearest neighbor (KNN) models. The KNN model was substantially easier to implement than the RF approach but both were highly accurate in classifying basic behaviors such as flapping (85.5% and 83.6% accurate, respectively), soaring (92.8% and 87.6%) and sitting (84.1% and 88.9%) with overall accuracies of 86.6% and 92.3% respectively. More detailed classification schemes, with specific behaviors such as banking and straight flights were well classified only by the KNN model (91.24% accurate; RF = 61.64% accurate). The RF model maintained its accuracy of classifying basic behavior classification accuracy of basic behaviors at sampling frequencies as low as 10Hz, the KNN at sampling frequencies as low as 20Hz. Classification of accelerometer data collected from free ranging birds demonstrated a strong dependence of predicted behavior on the type of classification model used. Our analyses demonstrate the consequence of different approaches to classification of accelerometry data, the potential to optimize classification algorithms with validated flight behaviors to improve classification accuracy, ideal sampling frequencies for different classification algorithms, and a number of ways to improve commonly used analytical techniques and best practices for classification of accelerometry data.

Identification of new spectral signatures associated with dengue virus infected sera

In a quest to use Raman spectroscopy as an optical diagnostic tool, we recorded Raman spectra of 32 dengue virus (DENV)‐infected and 28 healthy sera samples in the near‐infrared spectral range (540 to 1700 cm−1) using laser at 785 nm as the excitation source. We observed clear differences in the Raman spectra of DENV‐infected sera as compared with those of healthy individuals. Here, as a result of our study, we report 12 unique Raman bands associated with DENV‐infected sera that are not reported earlier. After applying analysis of variance and t‐test (p < 0.05) on these 12 bands, six Raman bands at 630 (N‐acetylglucosamine), 883 (in‐plane bending (ring) of deoxyribose), 1218 (amide III–β conformation from C6H5–C stretching vibrations of tryptophan and phenylalanine), 1273 (amide–III), 1623 (tryptophan) and 1672 cm−1 (ceramide) were found only in the DENV‐infected sera. The remaining six Raman bands at 716 (lipids), 780 (Uracil‐based ring breathing mode), 828 (ring breathing tyrosine), 840 (α‐anomers), 1101 (ν(C–N) of lipids and DNA) and 1150 cm−1(glycogen/carotenoids) were only found in healthy sera. Two types of classification models, principal component analysis and linear discriminant analysis, were employed to develop principal component analysis–linear discriminant analysis model that has provided diagnostic accuracy 96.50%, sensitivity 93.44%, and specificity 100%. This indicates that these 12 Raman bands have the potential to be used as biomarkers for optical diagnosis of DENV infection. This study provides a new insight for future research in the field of optical diagnosis using Raman spectroscopy. Copyright © 2017 John Wiley & Sons, Ltd.

Prediction of Bankruptcy with SVM Classifiers Among Retail Business Companies in EU

Article focuses on the prediction of bankruptcy of the 850 medium-sized retail business companies in EU from which 48 companies gone bankrupt in 2014 with respect to lag of the used features. From various types of classification models we chose Support vector machines method with linear, polynomial and radial kernels to acquire best results. Pre-processing is enhanced with filter based feature selection like Gain ratio, Chi-square and Relief algorithm to acquire attributes with the best information value. On this basis we deal with random samples of financial data to measure prediction accuracy with the confusion matrices and area under curve values for different kernel types and selected features. From the results it is obvious that with the rising distance to the bankruptcy there drops precision of bankruptcy prediction. The last year (2013) with avaible financial data offers best total prediction accuracy, thus we also infer both the Error I and II types for better recognizance. The 3rd order polynomial kernel offers better accuracy for bankruptcy prediction than linear and radial versions. But in terms of the total accuracy we recommend to use radial kernel without feature selection.