摘要： 对Ｗｅｂ网页识别中有关特征选择的两个重要问题进行了较为深入的探讨．提出了一种新的描述特征选择方法，并将其与３种已有的描述特征选择方法进行实验比较，证实其有效性；此外还对５种在文本归类中，具有代表性的识别特征选择方法在Ｗｅｂ网页识别中的实际应用效果进行了评估比较，并发现信息增益和统计方法，选择识别特征效果最佳．

关键词: 特权选择, 网页分类, 决策树, 机器学习

Abstract: This paper studies two related important problems in the Web page classification deeply. With regard to the description feature selection problem. a new method is put forward and compared with the other three existed methods to prove its effectiveness. Besides we have evaluated the practical effect of five learning feature selection methods used in the text categorization effectively before. and the information gain method and statistic method show better performance in the Web page classincation

中图分类号: 

• TP393