基于一种改进粒子群算法的SVM参数选取

支持向量机作为一个新兴的数学建模工具已经被广泛地应用到很多工业控制领域中,其良好的泛化能力和预测精度在很大程度上受到其参数选取的影响.根据智能群体进化模式改进粒子群优化算法.利用模糊C均值聚类算法分类粒子群体,并用子群体最优点取代速度更新公式中的个体历史最优点,并利用该算法搜索支持向量机的最优参数组合.对比仿真实验表明:所提优化算法是支持向量机参数选取的有效算法,在非线性函数估计中体现出优良的性能.

Parameters selection of SVM based on extended PSO algorithm

Support Vector Machine(SVM), a new mathematic modeling tool, has been widely used in many industry applications. The good generalization ability and estimation accuracy are impacted by parameters selection of SVM. The Particle Swarm Optimization (PSO) was improved based on evolution model of intelligence group. The whole group was divided into small groups by fuzzy C-means clustering algorithm. The individual best points in velocity updating function were replaced by the best points in small groups. At last, the extended PSO algorithm was proposed to search the optimal combination of SVM parameters. Simulations show that the proposed algorithm is an effective way to search the SVM parameters and has good performance in nonlinear function estimation.