L2-loss twin support vector machine for classification

Abstract

Twin support vector machine (TSVM) is a rapid algorithm for resolving discriminating problems using a pair of quadratic programming problems (QPPs). Based on the TSVM and SVM, this paper proposes regularization twin support vector machine with L2 loss function (L2-RTSVM) for Classification, the coordinate descent algorithm with shrinking technique is used to solve the L2-RTSVM. L2-RTSVM has higher classification accuracy and efficiency than TSVM, and overcomes the drawback of TSVM. The experiments show that the performance of L2-RTSVM is better than those of SVM, TSVM and TPMSVM in accuracy and time.