Abstract:

To solve the overfitting problems with support vector machine (SVM) for the outlier or noise, the characteristics of fuzzy support vector machine (FSVM) and proximal support vector machine (PSVM) are analyzed. Drawn on their advantages, namely, fuzzy membership and proximal hyper plane, a method based on support vector data domain (SVDD) in description is proposed. This method fully considers the relationship between the distances of each sample point to the center of each class and the contribution rate of each sample. The improved algorithm performs more clearly and precise. The analytical results show the algorithm with fuzzy membership degree has a higher recognition rate, but spents a greater amount of training time.