Granular twin support vector machines based on mixture kernel function

Abstract

The recently proposed twin support vector machines, denoted by TWSVM, gets perfect classification performance and is suitable for many cases. However, it would reduce its learning performance when it is used to solve the large number of samples. In order to solve this problem, a novel algorithm called Granular Twin Support Vector Machines based on Mixture Kernel Function (GTWSVM-MK) is proposed. Firstly, a grain method including coarse particles and fine particles is propsed and then the judgment and extraction methods of support vector particles are given. On the above basis, we propose a granular twin support vector machine learning model. Secondly, in order to solve the kernel function selection problem, minxture kernel function is introduced. Finally, compared with SVM and TWSVM, the experimental results show that GTWSVM-MK has higher classification performance.