一种基于类别的组合型文本特征选择  

A combined text feature selection algorithm based on category information

作　　者：刘海峰[1] 赵华[1] 刘守生[1] 

机构地区：[1]解放军理工大学理学院,南京210007

出　　处：《情报学报》2010年第4期744-748,共5页Journal of the China Society for Scientific and Technical Information

基　　金：国家自然科学基金资助项目（编号：70571087）

摘　　要：文本特征降维对文本分类的精确性有着非常重要的影响。本文针对传统的TF-IDF没有考虑特征项的类间分布状况以及对类属低频词的抑制现象、MI在训练样本类别分布偏斜条件下的不足问题分别进行了改进,进而提出了一种基于类别的组合型文本特征选择算法。随后的文本分类试验表明,本文提出的加权模型相比较于传统的TF-IDF以及MI方法可以有效提高文本分类的精度。The quality of text feature reduction affects the accuracy of text categorization.Due to the deficiency of traditional TF-IDF without considering the distribution of feature words among classes and the small frequency sort words have been restrain, and more the shortage of MI without considering the text class tilt,the paper improves these disadvantage.Basing on this,the paper proposes a combined text feature selection algorithm based on category information.The test about text categorization shows that this method is valid in improving the accuracy of text categorization.

关 键 词：文本分类 特征选择 互信息方法 特征加权 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]