Search results

Ensemble models that combine multiple base classifiers have been widely used to improve prediction performance in credit risk evaluation. However, an arbitrary selection of base classifiers is problematic. The purpose of this paper is to develop a framework for selecting base classifiers to improve the overall classification performance of an ensemble model.

In this study, selecting base classifiers is treated as a feature selection problem, where the output from a base classifier can be considered a feature. The proposed correlation-based classifier selection using the maximum information coefficient (MIC-CCS), a correlation-based classifier selection under the maximum information coefficient method, selects the features (classifiers) using nonlinear optimization programming, which seeks to optimize the relationship between the accuracy and diversity of base classifiers, based on MIC.

The empirical results show that ensemble models perform better than stand-alone ones, whereas the ensemble model based on MIC-CCS outperforms the ensemble models with unselected base classifiers and other ensemble models based on traditional forward and backward selection methods. Additionally, the classification performance of the ensemble model in which correlation is measured with MIC is better than that measured with the Pearson correlation coefficient.

The study provides an alternate solution to effectively select base classifiers that are significantly different, so that they can provide complementary information and, as these selected classifiers have good predictive capabilities, the classification performance of the ensemble model is improved.

This paper introduces MIC to the correlation-based selection process to better capture nonlinear and nonfunctional relationships in a complex credit data structure and construct a novel nonlinear programming model for base classifiers selection that has not been used in other studies.

The aim of this study is to investigate subject independent emotion recognition capabilities of EEG and peripheral physiological signals namely: electroocoulogram (EOG), electromyography (EMG), electrodermal activity (EDA), temperature, plethysmograph and respiration. The experiments are conducted on both modalities independently and in combination. This study arranges the physiological signals in order based on the prediction accuracy obtained on test data using time and frequency domain features.

DEAP dataset is used in this experiment. Time and frequency domain features of EEG and physiological signals are extracted, followed by correlation-based feature selection. Classifiers namely – Naïve Bayes, logistic regression, linear discriminant analysis, quadratic discriminant analysis, logit boost and stacking are trained on the selected features. Based on the performance of the classifiers on the test set, the best modality for each dimension of emotion is identified.

 The experimental results with EEG as one modality and all physiological signals as another modality indicate that EEG signals are better at arousal prediction compared to physiological signals by 7.18%, while physiological signals are better at valence prediction compared to EEG signals by 3.51%. The valence prediction accuracy of EOG is superior to zygomaticus electromyography (zEMG) and EDA by 1.75% at the cost of higher number of electrodes. This paper concludes that valence can be measured from the eyes (EOG) while arousal can be measured from the changes in blood volume (plethysmograph). The sorted order of physiological signals based on arousal prediction accuracy is plethysmograph, EOG (hEOG + vEOG), vEOG, hEOG, zEMG, tEMG, temperature, EMG (tEMG + zEMG), respiration, EDA, while based on valence prediction accuracy the sorted order is EOG (hEOG + vEOG), EDA, zEMG, hEOG, respiration, tEMG, vEOG, EMG (tEMG + zEMG), temperature and plethysmograph.

Many of the emotion recognition studies in literature are subject dependent and the limited subject independent emotion recognition studies in the literature report an average of leave one subject out (LOSO) validation result as accuracy. The work reported in this paper sets the baseline for subject independent emotion recognition using DEAP dataset by clearly specifying the subjects used in training and test set. In addition, this work specifies the cut-off score used to classify the scale as low or high in arousal and valence dimensions. Generally, statistical features are used for emotion recognition using physiological signals as a modality, whereas in this work, time and frequency domain features of physiological signals and EEG are used. This paper concludes that valence can be identified from EOG while arousal can be predicted from plethysmograph.

Ensemble methods have been widely used in the field of pattern recognition due to the difficulty of finding a single classifier that performs well on a wide variety of problems. Despite the effectiveness of these techniques, studies have shown that ensemble methods generate a large number of hypotheses and that contain redundant classifiers in most cases. Several works proposed in the state of the art attempt to reduce all hypotheses without affecting performance.

In this work, the authors are proposing a pruning method that takes into consideration the correlation between classifiers/classes and each classifier with the rest of the set. The authors have used the random forest algorithm as trees-based ensemble classifiers and the pruning was made by a technique inspired by the CFS (correlation feature selection) algorithm.

The proposed method CES (correlation-based Ensemble Selection) was evaluated on ten datasets from the UCI machine learning repository, and the performances were compared to six ensemble pruning techniques. The results showed that our proposed pruning method selects a small ensemble in a smaller amount of time while improving classification rates compared to the state-of-the-art methods.

CES is a new ordering-based method that uses the CFS algorithm. CES selects, in a short time, a small sub-ensemble that outperforms results obtained from the whole forest and the other state-of-the-art techniques used in this study.

During the production process of steel strip, some defects may appear on the surface, that is, traditional manual inspection could not meet the requirements of low-cost and high-efficiency production. The purpose of this paper is to propose a method of feature selection based on filter methods combined with hidden Bayesian classifier for improving the efficiency of defect recognition and reduce the complexity of calculation. The method can select the optimal hybrid model for realizing the accurate classification of steel strip surface defects.

A large image feature set was initially obtained based on the discrete wavelet transform feature extraction method. Three feature selection methods (including correlation-based feature selection, consistency subset evaluator [CSE] and information gain) were then used to optimize the feature space. Parameters for the feature selection methods were based on the classification accuracy results of hidden Naive Bayes (HNB) algorithm. The selected feature subset was then applied to the traditional NB classifier and leading extended NB classifiers.

The experimental results demonstrated that the HNB model combined with feature selection approaches has better classification performance than other models of defect recognition. Among the results of this study, the proposed hybrid model of CSE + HNB is the most robust and effective and of highest classification accuracy in identifying the optimal subset of the surface defect database.

The main contribution of this paper is the development of a hybrid model combining feature selection and multi-class classification algorithms for steel strip surface inspection. The proposed hybrid model is primarily robust and effective for steel strip surface inspection.

This paper aims to deal with the previously unknown prediction accuracy of students’ activity pattern in a blended learning environment.

To extract the most relevant activity feature subset, different feature-selection methods were applied. For different cardinality subsets, classification models were used in the comparison.

Experimental evaluation oppose the hypothesis that feature vector dimensionality reduction leads to prediction accuracy increasing.

Improving prediction accuracy in a described learning environment was based on applying synthetic minority oversampling technique, which had affected results on correlation-based feature-selection method.

The major contribution of the research is the proposed methodology for selecting the optimal low-cardinal subset of students’ activities and significant prediction accuracy improvement in a blended learning environment.

Software fault prediction is an important concept that can be applied at an early stage of the software life cycle. Effective prediction of faults may improve the reliability and testability of software systems. As service-oriented architecture (SOA)-based systems become more and more complex, the interaction between participating services increases frequently. The component services may generate enormous reports and fault information. Although considerable research has stressed on developing fault-proneness prediction models in service-oriented systems (SOS) using machine learning (ML) techniques, there has been little work on assessing how effective the source code metrics are for fault prediction. The paper aims to discuss this issue.

In this paper, the authors have proposed a fault prediction framework to investigate fault prediction in SOS using metrics of web services. The effectiveness of the model has been explored by applying six ML techniques, namely, Naïve Bayes, Artificial Networks (ANN), Adaptive Boosting (AdaBoost), decision tree, Random Forests and Support Vector Machine (SVM), along with five feature selection techniques to extract the essential metrics. The authors have explored accuracy, precision, recall, f-measure and receiver operating characteristic curves of the area under curve values as performance measures.

The experimental results show that the proposed system can classify the fault-proneness of web services, whether the service is faulty or non-faulty, as a binary-valued output automatically and effectively.

One possible threat to internal validity in the study is the unknown effects of undiscovered faults. Specifically, the authors have injected possible faults into the classes using Java C3.0 tool and only fixed faults are injected into the classes. However, considering the Java C3.0 community of development, testing and use, the authors can generalize that the undiscovered faults should be few and have less impact on the results presented in this study, and that the results may be limited to the investigated complexity metrics and the used ML techniques.

In the literature, only few studies have been observed to directly concentrate on metrics-based fault-proneness prediction of SOS using ML techniques. However, most of the contributions are regarding the fault prediction of the general systems rather than SOS. A majority of them have considered reliability, changeability, maintainability using a logging/history-based approach and mathematical modeling rather than fault prediction in SOS using metrics. Thus, the authors have extended the above contributions further by applying supervised ML techniques over web services metrics and measured their capability by employing fault injection methods.

Indian Premier League (IPL) is one of the more popular cricket world tournaments, and its financial is increasing each season, its viewership has increased markedly and the betting market for IPL is growing significantly every year. With cricket being a very dynamic game, bettors and bookies are incentivised to bet on the match results because it is a game that changes ball-by-ball. This paper investigates machine learning technology to deal with the problem of predicting cricket match results based on historical match data of the IPL. Influential features of the dataset have been identified using filter-based methods including Correlation-based Feature Selection, Information Gain (IG), ReliefF and Wrapper. More importantly, machine learning techniques including Naïve Bayes, Random Forest, K-Nearest Neighbour (KNN) and Model Trees (classification via regression) have been adopted to generate predictive models from distinctive feature sets derived by the filter-based methods. Two featured subsets were formulated, one based on home team advantage and other based on Toss decision. Selected machine learning techniques were applied on both feature sets to determine a predictive model. Experimental tests show that tree-based models particularly Random Forest performed better in terms of accuracy, precision and recall metrics when compared to probabilistic and statistical models. However, on the Toss featured subset, none of the considered machine learning algorithms performed well in producing accurate predictive models.

The authorship authentication (AA) problem is concerned with correctly attributing a text document to its corresponding author. Historically, this problem has been the focus of various studies focusing on the intuitive idea that each author has a unique style that can be captured using stylometric features (SF). Another approach to this problem, known as the bag-of-words (BOW) approach, uses keywords occurrences/frequencies in each document to identify its author. Unlike the first one, this approach is more language-independent. This paper aims to study and compare both approaches focusing on the Arabic language which is still largely understudied despite its importance.

Being a supervised learning problem, the authors start by collecting a very large data set of Arabic documents to be used for training and testing purposes. For the SF approach, they compute hundreds of SF, whereas, for the BOW approach, the popular term frequency-inverse document frequency technique is used. Both approaches are compared under various settings.

The results show that the SF approach, which is much cheaper to train, can generate more accurate results under most settings.

Numerous advantages of efficiently solving the AA problem are obtained in different fields of academia as well as the industry including literature, security, forensics, electronic markets and trading, etc. Another practical implication of this work is the public release of its sources. Specifically, some of the SF can be very useful for other problems such as sentiment analysis.

This is the first study of its kind to compare the SF and BOW approaches for authorship analysis of Arabic articles. Moreover, many of the computed SF are novel, while other features are inspired by the literature. As SF are language-dependent and most existing papers focus on English, extra effort must be invested to adapt such features to Arabic text.

The purpose of this paper is to identify electronic word-of-mouth (eWOM) customers from customer reviews. Thus, firms can precisely leverage eWOM customers to increase their product sales.

This research proposed a framework to analyze the content of consumer-generated product reviews. Specific algorithms were used to identify potential eWOM reviewers, and then an evaluation method was used to validate the relationship between product sales and the eWOM reviewers identified by the authors’ proposed method.

The results corroborate that online product reviews that are made by the eWOM customers identified by the authors’ proposed method are more related to product sales than customer reviews that are made by non-eWOM customers and that the predictive power of the reviews generated by eWOM customers are significantly higher than the reviews generated by non-eWOM customers.

The proposed method is useful in the data set, which is based on one type of products. However, for other products, the validity must be tested. Previous eWOM customers may have no significant influence on product sales in the future. Therefore, the proposed method should be tested in the new market environment.

By combining the method with the previous customer segmentation method, a new framework of customer segmentation is proposed to help firms understand customers’ value specifically.

This study is the first to identify eWOM customers from online reviews and to evaluate the relationship between reviewers and product sales.

Breast cancer (BC) is one of the leading cancer in the world, BC risk has been there for women of the middle age also, it is the malignant tumor. However, identifying BC in the early stage will save most of the women’s life. As there is an advancement in the technology research used Machine Learning (ML) algorithm Random Forest for ranking the feature, Support Vector Machine (SVM), and Naïve Bayes (NB) supervised classifiers for selection of best optimized features and prediction of BC accuracy. The estimation of prediction accuracy has been done by using the dataset Wisconsin Breast Cancer Data from University of California Irvine (UCI) ML repository. To perform all these operation, Anaconda one of the open source distribution of Python has been used. The proposed work resulted in extemporize improvement in the NB and SVM classifier accuracy. The performance evaluation of the proposed model is estimated by using classification accuracy, confusion matrix, mean, standard deviation, variance, and root mean-squared error.

The experimental results shows that 70-30 data split will result in best accuracy. SVM acts as a feature optimizer of 12 best features with the result of 97.66% accuracy and improvement of 1.17% after feature reduction. NB results with feature optimizer 17 of best features with the result of 96.49% accuracy and improvement of 1.17% after feature reduction.

The study shows that proposal model works very effectively as compare to the existing models with respect to accuracy measures.