Dual Membership Fuzzy Support Vector Machine Algorithm Based on SVDD

Abstract

In the case of excessive overlap between positive and negative samples in data set, the deviation in the category of reconstructed sample points will lead to unsatisfactory discrimination of SVM, no matter what methods are used to reconstruct the sample set. A dual membership fuzzy support vector machine algorithm based on support vector data domain description was thus proposed, followed by a simulation analysis of common data set. Experimental results show that the proposed algorithm can work well in classification when the sample set is overlapped.