Text Classification Using ES Based L1-LS-SVM

Abstract

With the advent of big-data age, it is essential to organize, analyze, retrieve and protect the useful data or sensitive information in a fast and efficient way for customers from different industries and fields. In this paper, evolution strategies based a least squares support vector machine with L1 penalty (ES based L1-LS-SVM) is proposed to deal with LS-SVM shortcomings. A minimum of 1-norm based object function is chosen to get the sparse and robust solution based on the idea of basis pursuit (BP) in the whole feasibility region. A real Chinese corpus from Fudan University is used to demonstrate the effectiveness of this model. The experimental results show that ES based L1-LS-SVM can obtain a small number of support vectors and improve the generalization ability of ES based LS-SVM.