A Double Margin Based Fuzzy Support Vector Machine Algorithm

Abstract

Although fuzzy support vector machine introduces the fuzzy membership degree in maximizing the margin and improves performance of classifier, it has not fully considered the position of training samples in the margin. In this paper, a double margin (rough margin) based fuzzy support vector machine (RFSVM) algorithm is presented by introducing rough set into fuzzy support vector machine. Firstly, we compute the degree of fuzzy membership of each training sample. Secondly, the data with fuzzy memberships are trained to obtain the decision hyperplane that maximizing rough margin method which contains the lower margin and the upper margin. In this algorithm, points in the lower margin have major penalty than those in the boundary in the rough margin. Finally, experiments on several benchmark datasets show that the RFSVM algorithm is very effective and feasible relative to the existing support vector machines.