The parameters selection for SVM based on improved chaos optimization algorithm

Abstract

The parameters selection of support vector machine decides its study performance and generalization ability. The SVM model is greatly influenced by penalty factor and the kernel function parameter such as σ for the radial basis function (RBF) kernel. To searching the best compound of parameters, a new algorithm is proposed based on improved chaos optimization strategy to realized automatic parameters selection for SVM. Chaos optimization algorithm is a global searching method in which the complexity and dimension of variables need not to be considered. Compared with the algorithms based on GA and PSO, the classification efficiency is improved greatly. © 2011 Springer-Verlag.