Computer Engineering and Applications ›› 2010, Vol. 46 ›› Issue (26): 142-144.DOI: 10.3778/j.issn.1002-8331.2010.26.044

• 数据库、信号与信息处理 • Previous Articles     Next Articles

Research of improving multicollinearity in SVM classification algorithm

XIAN Guang-ming1,2,QI De-yu1,FANG Qun2,KE Qing3,ZENG Bi-qing4,XIAO Ying-wang4   

  1. 1.South China University of Technology,Guangzhou 510640,China
    2.Guangzhou Tianhe Software Park MC post-Doctoral Research Station,Guangzhou 510663,China
    3.Bluedon Information Security Technology Co.,Ltd,Guangzhou 510631,China
    4.Nanhai Campus,South China Normal University,Foshan,Guangdong 528225,China
  • Received:2008-04-07 Revised:2009-12-04 Online:2010-09-11 Published:2010-09-11
  • Contact: XIAN Guang-ming

改进SVM分类算法中多重共线性问题研究

冼广铭1,2,齐德昱1,方 群2,柯 庆3,曾碧卿4,肖应旺4   

  1. 1.华南理工大学,广州 510640
    2.广州天河软件园管委会博士后科研工作站,广州 510663
    3.蓝盾信息安全技术股份有限公司,广州 510631
    4.华南师范大学 南海学院,广东 佛山 528225
  • 通讯作者: 冼广铭

Abstract: In order to solve this limitation problem of multicollinearity in SVM classification algorithm,the method of factor analysis is proposed.Factor analysis is used to represent most information of original variables by several independent factors.It greatly decreases number of variable for data construction,simplifies construction of SVM,and decreases complexity of classification process of SVM.At the same time,sample distribution characters are remained.Experimental results show that the method above can effectively solve the problem of multicollinearity in SVM classification algorithm.

Key words: support vector machine, factor analysis, multicollinearity

摘要: 提出了一种可以解决SVM分类算法中的多重共线性问题的因子分析方法。因子分析的核心是用较少的互相独立的因子反映原有变量的绝大部分信息,它既能大大减少参与数据建模的变量个数,简化支持向量机结构,减少支持向量机分类过程中的复杂度和运算量,同时不会改变样本的分布特性,保持样本的分类信息。实验结果表明,通过因子分析对样本数据的处理,使用3个因子代替7个原始变量,原始变量间的多重共线性问题得到了很好的解决。

关键词: 支持向量机, 因子分析, 多重共线性

CLC Number: