Multi-category news classification using Support Vector Machine based classifiers

Support Vector Machine (SVM) and its variants are gaining momentum among the Machine Learning community. In this paper, we present a quantitative analysis between the established SVM based classifiers on multi-category text classification problem. Here, we are particularly interested in studying the behaviour of Least-squares Support Vector Machines, Twin Support Vector Machines and Least-squares Twin Support Vector Machines (LS-TWSVM) classifiers on News data. Since, all these are binary classifiers, they are extended using One-Against-All approach to handle multi-category data. The dataset is first converted into required format by performing preprocessing activities which involve tokenization and removing irrelevant data. The feature set is constructed as Term Frequency-Inverse Document Frequency matrix, so that representative vectors could be obtained for each document. Experimentally, we have compared the performance of each classification algorithm by performing simulations on benchmark UCI News datasets: Reuters and 20 Newsgroups. This paper shows that LS-TWSVM proves to be the best of all three, both in terms of accuracy and time complexity (training and testing).