基于CFOA的混合核SVM参数优化

Parameter optimization and application of SVM with mixture kernels based on CFOA

针对混合核函数支持向量机(SVM)的多参数选择问题,利用具有较强全局搜索能力的混沌果蝇优化算法(CFOA)对混合核函数SVM中的重要参数进行优化调整。引入基于Lozi’s映射的混沌算法,提高果蝇种群的多样性和搜索的遍历性,有效避免局部最优;在果蝇优化算法中使用负线性搜索距离,提高算法精度。利用UCI数据库进行测试,测试结果表明,CFOA算法能够快速有效地提取混合核SVM的最佳参数组合,分类效果更好。

To solve the multi-parameter optimization problem of SVM with mixture kernels,a chaotic fruit fly optimization algorithm（CFOA）with global searching ability was proposed to find the best combination of the basic parameters.By introducing the chaos algorithm based on Lozi＇s mapping into the evolutionary process of basic FOA,the performances of the algorithm including population diversity and fruit searching ergodicity were improved,thus effectively avoiding local extremum.The use of negative linear searching distance in fruit fly optimization improved the accuracy of the algorithm.Simulations on UCI data show that the proposed algorithm provides an effective way to search the best parameters combination,and makes SVM have better performance and classification accuracy.