An improved fuzzy support vector machine algorithm based on refactoring class center

作者：Zhou, Huaping^*; Qin, Huangli; Shen, Huali

来源：International Conference on Applications and Techniques in Cyber Intelligence, ATCI 2018, Shanghai, China, 2018-07-11 to 2018-07-13.

DOI：10.1007/978-3-319-98776-7_11

• Class Centers
• Classification accuracy
• Degree of membership
• Essential characteristic
• Function designs
• Fuzzy support vector machines
• Geometric distribution
• Membership values

The fuzzy support vector machine with membership function designed based on class center can effectively solve the problems of traditional support vector machine (SVM in sensitivity to noises and outliners. However it reduces the classification effect by endow support vectors with smaller membership values. Therefore, the paper proposed an improved membership function design method, which reduces the dependence on geometric distribution of the sample data while considering the essential characteristic of SVM as well. To ensure new class center closer to the surface of classification. A new class center is calculated according to the average value of the distance between the hyperplane and the data which have major impact on classification accuracy. Thus improved the classification accuracy by samples which are closer to the classification of surface of classification obtained the higher degree of membership and others get lower degree of membership.

• 出版日期2019
• 单位安徽大学