计算机工程与应用 ›› 2011, Vol. 47 ›› Issue (29): 137-139.

• 数据库、信号与信息处理 • 上一篇    下一篇

基于分布距离的特征聚类方法

张学谦,王自强,郜凤敏   

  1. 河南工业大学 信息科学与工程学院,郑州 450001
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2011-10-11 发布日期:2011-10-11

Feature clustering method based on distribution distance

ZHANG Xueqian,WANG Ziqiang,GAO Fengmin   

  1. College of Information Science and Engineering,Henan University of Technology,Zhengzhou 450001,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2011-10-11 Published:2011-10-11

摘要: 为降低特征空间维数,提出了一种基于分布距离的文本特征聚类方法,通过将特征空间中分布距离相近的特征聚合,来实现降维。在TanCorpusV1.0语料库上实验表明,当将特征空间维数降低至原空间的近10%时,用SVM作为分类器,获得了比特征提取方法高的分类精度。

关键词: 分布特征, 分布距离, 特征抽取, 特征聚类

Abstract: To reduce feature space dimensionality,this paper presents a new method to cluster the similar features based on distribution distance,which can achieve dimensionality reduction through clustering the nearest distance features.Test on the corpus of TanCorpusV1.0 shows,when reducing the dimensionality of feature space as far as original’s 10%,using SVM as classifier,this method can achieve a higher accuracy than feature selection method.

Key words: distribution feature, distribution distance, feature extraction, feature clustering