A Fuzzy Classification Model with SVM

Abstract

A fuzzy classification model with support vector machine (FCMWSVM) is proposed. For the basic idea of constructing this model, firstly the kernel function is constructed by selecting suitable membership function. Then a fuzzy partition is built around each training pattern and a fuzzy IF-THEN classification rule is defined for each fuzzy partition. Finally, the support vectors and the parameters for rule are got by SVM learning method. The basic idea and the structure of this model are introduced. The effects of the membership function parameters and the penalty parameters for the classification rule and the classifier performance are analyzed. Experiments with two-spiral line data and typical data sets evaluate the performances of this model.