A new fuzzy support vector machine based on mixed kernel function

Support vector machine is effective method for resolving non-liner classification and regression problem, but it is sensitive to the noises and outliers in the training samples. In order to overcome this problem, fuzzy support vector machine (FSVM) is introduced. How to choose a proper fuzzy membership is very important for the practical problem in FSVM. Generally, fuzzy membership is built according to the distance between each input date point and its class center in primal space. In this paper, a new fuzzy membership function is proposed to construct using mixed kernel function in future space. The experiments show that its superiority comparing with traditional SVM and FSVM, and conventional kernel and mixed kernel.