计算机工程与应用 ›› 2009, Vol. 45 ›› Issue (24): 143-145.DOI: 10.3778/j.issn.1002-8331.2009.24.042

• 数据库、信息处理 • 上一篇    下一篇

支持向量机算法多目标模型选择

黄景涛,池小梅,马建伟   

  1. 河南科技大学 电子信息工程学院,河南 洛阳 471003
  • 收稿日期:2009-04-07 修回日期:2009-06-15 出版日期:2009-08-21 发布日期:2009-08-21
  • 通讯作者: 黄景涛

Multi-object model selection of Support Vector Machine

HUANG Jing-tao,CHI Xiao-mei,MA Jian-wei   

  1. Electronic & Information Engineering College,Henan University of Science and Technology,Luoyang,Henan 471003,China
  • Received:2009-04-07 Revised:2009-06-15 Online:2009-08-21 Published:2009-08-21
  • Contact: HUANG Jing-tao

摘要: 为适应支持向量机(Support Vector Machine,SVM)算法应用过程中的不同性能指标要求,将SVM算法的模型选择问题作为一个多目标优化(Multi-Object Optimization,MOO)问题进行处理。以改进的粒子群优化(Particle Swarm Optimization,PSO)算法对该多目标优化问题进行求解,得到其Pareto解集,在具体应用中根据实际需要从Pareto解集中选择适合的最优解作为支持向量机算法参数,实现支持向量机算法的模型选择。在几个数据集上的仿真实验表明,该方法能够较快地得到Pareto解集,解集中的参数组合能够满足对支持向量机算法速度和泛化能力的不同要求。

关键词: 支持向量机(SVM), 模型选择, 多目标优化(MOO), 粒子群优化(PSO)

Abstract: To meet the different focuses on the performances of Support Vector Machine(SVM) during application,the model selection of SVM is taken as a Multi-Object Optimization(MOO) problem.Particle Swarm Optimization(PSO) is applied to solve this MOO problem.The solution known as Pareto front is gained and one can select a concrete single solution from it according to own application needs,i.e.final model selection.The experiments on several datasets show that the method can gain Pareto front faster and the elements in the Pareto set can meet the need on generalization performance and training speed in SVM application.

Key words: Support Vector Machine(SVM), model selection, Multi-Object Optimization(MOO), Particle Swarm Optimization(PSO)

中图分类号: