自动文本分类特征选择方法研究

文本分类是指根据文本的内容将大量的文本归到一个或多个类别的过程，文本表示技术是文本分类的核心技术之一，而特征选择又是文本表示技术的关键技术之一，对分类效果至关重要。文本特征选择是最大程度地识别和去除冗余信息，提高训练数据集质量的过程。对文本分类的特征选择方法，包括信息增益、互信息、X^2统计量、文档频率、低损降维和频率差法等做了详细介绍、分析、比较研究。

Automatic text categorization feature selection methods research

Text categorization is a task of classifying text documents into predefined set of categories based on their content, text representation is one of kernel technology of text categorization, and feature selection is one of key technology of text representation, it is very important to text categorization effect. Text feature selection is a process of recognizing and deleting redundant information and enhancing training documents cluster quality. The text feature selection methods are introduced, analysed and researched, including information gain, mutual information, X^2 statistics, document frequency, low loss dimensionality reduction, relative frequency difference.