使用粒子群算法进行特征选择及对支持向量机参数的优化

特征选取和参数设置是提升支持向量机分类器的效果的两个主要手段.为了将两者结合起来,实现同步优化,以达到更好的分类效果,设计了一种基于粒子群算法的分类器优化算法.新算法对粒子采用2进制编码的,设计适合的目标函数,同步进行特征选择和支持向量机参数的优化.经过对比验证,新方法能够更加准确的得到待分类数据的特征子集跟支持向量机参数,最终得到更优的处理结果.

Use PSO to Perform Feature Selection and Parameter Optimization of SVM

Feature selection as while as parameter optimization are two important way to improve SVM performance.In order to combine the two different way together to get a better classification effect,an new optimization algorithm based PSO is designed.The new algorithm uses Binary encode,and constructs proper target function to optimize the Feature selection and parameter optimization at the same time.Experiment s show that the new algorithm can quickly and exactly get a better suitable feature subset s and SVM parameters,and get a better classification effect.