计算机工程与应用 ›› 2010, Vol. 46 ›› Issue (30): 134-137.DOI: 10.3778/j.issn.1002-8331.2010.30.040

• 数据库、信号与信息处理 • 上一篇    下一篇

使用类内集中度和分层递阶约简的特征选择方法

陈吕强1,朱颢东2,3,伏明兰1   

  1. 1.黄山学院 信息工程学院,安徽 黄山 245021
    2.郑州轻工业学院 计算机与通信工程学院,郑州 450002
    3.中国科学院 成都计算机应用研究所,成都 610041
  • 收稿日期:2009-03-26 修回日期:2010-03-18 出版日期:2010-10-21 发布日期:2010-10-21
  • 通讯作者: 陈吕强

Feature selection method using classificatory concentration and hierarchical reduction

CHEN Lv-qiang1,ZHU Hao-dong2,3,FU Ming-lan1   

  1. 1.Department of Information Engineering,Huangshan University,Huangshan,Anhui 245021,China
    2.College of Computer and Communication Engineering,Zhengzhou University of Light Industry,Zhengzhou 450002,China
    3.Chengdu Institute of Computer Application,Chinese Academy of Sciences,Chengdu 610041,China
  • Received:2009-03-26 Revised:2010-03-18 Online:2010-10-21 Published:2010-10-21
  • Contact: CHEN Lv-qiang

摘要: 特征选择是文本分类的关键步骤之一,所选特征子集的优劣直接影响文本分类的结果。首先简单分析了几种经典的特征选择方法,总结了它们的不足,然后提出了类内集中度的概念,紧接着把分层递阶的思想引入粗糙集并提出了一个改进的基于分层递阶的属性约简算法,最后把该约简算法同类内集中度结合起来,提出了一个综合的特征选择方法。该方法首先利用类内集中度进行特征初选以过滤掉一些词条来降低特征空间的稀疏性,然后利用所提约简算法消除冗余,从而获得较具代表性的特征子集。实验结果表明此种特征选择方法效果良好。

关键词: 文本分类, 特征选择, 类内集中度, 分层递阶约简

Abstract: Feature selection is one of the key steps in text categorization.The selected feature subset directly influences results of text categorization.It firstly analyzes simply several classic feature selection methods and summarizes their deficiencies,and then presents the concept of classificatory concentration,next introduces thoughts of hierarchical reduction into rough sets and provides an improved hierarchical reduction algorithm,finally,combines the reduction algorithm with the classificatory concentration and proposes a comprehensive feature selection method.The comprehensive method firstly uses the classificatory concentration to select feature and filter out some terms to reduce the sparsity of feature spaces,and then uses the hierarchical reduction algorithm to eliminate redundancy,so can acquire the feature subset which are more representative.The experimental results show that the comprehensive method is promising.

Key words: text categorization, feature selection, classificatory concentration, hierarchical reduction

中图分类号: