A Genetic Algorithm Approach to Kernel Functions Parameters Selection for SVM

Abstract

The Support Vector Machines (SVM) is a classification algorithm with many diverse applications. The SVM has many parameters associated with it which influences the performance of the SVM classifier. In this paper, we employ Genetic Algorithm based approach to find and select an appropriate kernel function and its parameters. This proposed technique combines predictive accuracy and complexity of SVM as two criteria into a fitness function for evaluating the performance of SVM. Our method is compared with grid algorithm and the experimental results validate that the proposed approach is much better than the grid method.