Weighted Twin Support Vector Machine with Universum

Abstract

Universum is a new concept proposed recently, which is defined to be the sample that does not belong to any classes concerned. Support Vector Machine with Universum (..-SVM) is a new algorithm, which can exploit Universum samples to improve the classification performance of SVM. In fact, samples in the different positions have different effects on the bound function. Then, we propose a weighted Twin Support Vector Machine with Universum (called ..-WTSVM), where samples in the different positions are proposed to give different penalties. Therefore ..-WTSVM has better flexibility of the algorithm and can obtain more reasonable classifier in most case. All experiments demonstrate that our ..-WTSVM far outperforms TSVM, and lightly outperforms ..-TSVM not only in linear case but also in nonlinear case.