Research of reducing text feature based on scatter difference criterion
基于散度差准则的文本特征降维研究*

The paper studied a method of extracting text feature based on scatter difference.Firstly,analyzed the primary feature reduction means and their characteristic in text classification. Secondly,analyzed the principle and its method that based on scatter difference criterion. And more,the paper demonstrated the validity about the main approaches of this method. Lastly,had a test about Chinese text categorization with this way.It improved the KNN classifier in the density factor to eliminate the data incline disadvantage. The result shows that this method has a better precision.