文本分类中特征提取方法的比较改进

摘要

Four kinds of commonly used text feature extraction method , namely, IG(Information gain), MI(Mu-tual information), CHI(χ2 statistical magnitude), DF(Document frequency) are introduced respectively.And then two improved methods are proposed according to the method of CHI , and analyzed by the experiment .It turns out to show that the improved method can increase the correct rate of text categorization .%介绍了IG（Information gain）信息增益，MI（Mutual information）互信息值，CHI（χ2统计法）、DF（ Document frequency ）文档频率4种常用的文本特征提取方法，然后针对CHI方法提出了改进方法，并对改进的方法进行了实验分析，结果表明改进的方法提高了文本分类的正确率。