• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

J4 ›› 2011, Vol. 33 ›› Issue (10): 126-130.

• 论文 • 上一篇    下一篇

基于ACO算法的SVM核函数的参数优化

赵新建,沈友文,徐〓俊   

  1. (浙江工业大学计算机学院,浙江 杭州 310023)
  • 收稿日期:2010-09-09 修回日期:2010-12-20 出版日期:2011-10-25 发布日期:2011-10-25

Parameter Optimization of the SVM Kernel Function Based on the ACO Algorithm

ZHAO Xinjian,SHEN Youwen,XU Jun   

  1. (School of Computer Science and Technology,Zhejiang University of Technology,Hangzhou 310023,China)
  • Received:2010-09-09 Revised:2010-12-20 Online:2011-10-25 Published:2011-10-25

摘要:

本文针对核函数参数选择的随意性影响支持向量机分类性能的问题,提出了一种基于离散编码的蚁群算法(CCACODE)的SVM核函数优化模型。CCACODE解决了连续函数优化的蚁群算法(CACO)求解之前必须进行预处理的问题,解决了基于网格划分策略的连续域蚁群算法(CACOGT)在求解精度的缺点、最优解必在定义域内的等分割点问题。仿真结果验证了该方法的有效性, F1值达到了90%以上。

关键词: 蚁群算法, 核函数, 支持向量机, 参数优化

Abstract:

As to the problem that the arbitrariness selection of the kernel parameters affects the performance of SVM, this paper presents the SVM kernel function optimization model based on the discrete ACO algorithm (CCACODE). CCACODE solves the problem that the continuous ACO algorithm (CACO) must be pretreated; solves the precision shortcomings of the CACO based on a grid search strategy. The simulation result shows that the method is effective, and the  F1 value reaches more than 92%.

Key words: ACO;kernel function;svm;paramerter optimization