Parameters Selection of Hybrid Kernel Based on GA
Yingjian Meng · 2009
Support vector machines (SVMs), a powerful machine method proposed by Vapnik have made significant achievement in pattern classification and regression estimation. In practice, when we use the method of SVM there exist two problems: the selection of kernel and the selection of parameters. In this paper, we discuss the hybrid kernels and propose a new method to select parameters, which uses genetic algorithm by designing relevant fitness function. Using some benchmark data sets, we show the effectiveness of our method.