The support vector machine based on intuitionistic fuzzy number and kernel function

Fuzzy support vector machine applied a degree of membership to each training point and reformulated the traditional support vector machines, which reduced the effects of noises and outliers for classification. However, the degree of membership only considered the distance from samples to the class center in the sample space, while neglected the situation of samples in the feature space and easily mistook the edge support vectors as noises. To deal with the aforementioned problems, the support vector machine based on intuitionistic fuzzy number and kernel function is proposed. In the high-dimensional feature space, each training point is assigned with a corresponding intuitionistic fuzzy number by the use of kernel function. Then, a new score function of the intuitionistic fuzzy numbers is introduced to measure the contribution of each training point. In the end, the new support vector machine is constructed according to the score value of each training point. The simulation results demonstrate the effectiveness and superiority of the proposed method.