Support vector machine (SVM) considers all data points with the same importance in classification problems, therefore SVM is very sensitive to noisy data or outliers. Current fuzzy approach to two-class SVM introduces a fuzzy membership to each data point in order to reduce the sensitivity of less important data, however computing fuzzy memberships is still a challenge. It has been found that the performance of fuzzy SVM highly depends on the computation of fuzzy memberships, hence in this paper, we propose a new method to compute fuzzy memberships and we also extend the fuzzy approach for two-class SVM to one-class SVM. Experiments performed on a number of popular data sets to evaluation the proposed fuzzy SVMs show promising classification results.