An Improvement of Structural – Twin Support Vector Machine

Abstract

In binary classification problems, two classes of data seem to be more complicated due to the number of data points of clusters in each class being different. Traditional algorithms such as Support Vector Machine (SVM), and Twin Support Vector Machine (TSVM) cannot sufficiently exploit structural information with cluster granularity. Structural Twin Support Vector Machine (STSVM) has exploited structural information with cluster granularity of data but does not use information about the number of data points in each cluster. This may affect the accuracy of classification problems. This paper proposes a new Improvement Structural - Support Vector Machine (called IS-SVM) for binary classification problems with a cluster-vs-class strategy. Experimental results show that the IS-SVM's training time is slower than that of TSVM and S-TSVM, but the IS-SVM's accuracy is better than that of TSVM and S-TSVM.