A New Fuzzy SVM based on the Posterior Probability Weighting Membership

To solve the sensitivity to the noises and outliers in support vector machine (SVM), the characterizations of fuzzy support vector machine (FSVM) are analyzed. But the determination of fuzzy membership is a difficulty. By the inspiration of bayesian decision theory and combining with sample density to give weight for each sample, new fuzzy membership function is proposed. Each sample points is given the tightness arranged forecasts by this method and the generalization ability of FSVM is improved. Numerical experiments show that, compared with the traditional SVM and FSVM, the improved algorithm performs, more effectively and accurately, has better classification result.