计算机工程与应用 ›› 2010, Vol. 46 ›› Issue (26): 122-124.DOI: 10.3778/j.issn.1002-8331.2010.26.038

• 数据库、信号与信息处理 • 上一篇    下一篇

优化的互信息特征选择方法

朱颢东1,2,3,陈 宁4,李红婵1   

  1. 1.郑州轻工业学院 计算机与通信工程学院,郑州 450002
    2.中国科学院 成都计算机应用研究所,成都 610041
    3.中国科学院 研究生院,北京 100039
    4.河南移动郑州分公司 网优中心,郑州 450052
  • 收稿日期:2010-04-12 修回日期:2010-07-23 出版日期:2010-09-11 发布日期:2010-09-11
  • 通讯作者: 朱颢东

Optimized mutual information feature selection method

ZHU Hao-dong1,2,3,CHEN Ning4,LI Hong-chan1   

  1. 1.School of Computer and Communication Engineering,Zhengzhou University of Light Industry,Zhengzhou 450002,China
    2.Chengdu Institute of Computer Applications,Chinese Academy of Sciences,Chengdu 610041,China
    3.The Graduate University of the Chinese Academy of Sciences,Beijing 100039,China
    4.The Net Optimal Center of Zhengzhou Branch of Henan Mobile,Zhengzhou 450052,China
  • Received:2010-04-12 Revised:2010-07-23 Online:2010-09-11 Published:2010-09-11
  • Contact: ZHU Hao-dong

摘要: 在文本分类中,互信息是一种被广泛应用的特征选择方法,但是该方法仅考虑了特征的文档频而没有考虑特征的词频,导致它经常倾向于选择出现频率较低的特征。为此,提出了一个新的文档频并把它引入到互信息方法中,从而获得了一种优化的互信息方法。该优化的互信息方法不但考虑了特征的文档频而且还考虑了特征出现的词频。实验结果表明该优化的互信息方法性能良好。

关键词: 文本分类, 互信息, 特征选择, 词频, 文档频

Abstract: Mutual Information(MI) is a feature selection method that is used widely in text categorization,but this method only takes into account document frequency of selected features and ignores word frequency of selected features.For this reason,MI is inclined to select lower-frequency features.In this case,a new document frequency is presented and introduced into MI,so that an optimized MI method is proposed.The optimized MI not only pays attention to document frequency of selected features but also attaches importance to word frequency of selected features.The experimental results show that the optimized MI is promising.

Key words: text categorization, Mutual Information(MI), feature selection, word frequency, document frequency

中图分类号: