Abstract: To improve the precision of search engine and locate user interesting Web page promptly, an investigation is made of how to provide directory navigation service in Tianwang, a spider style search engine The main idea is to implement Chinese Web page categorization automatically using supervised machine learning technology There are two important contributions in this paper: ①a large scale Chinese Web page data set in a hierarchy model is gathered and built, and it is the basis and premises to implement the Chinese Web page classification automatically; ②in consideration of the features of Chinese Web page and the inherent faults of CHI method, a new feature selection algorithm is implemented, which can reduce noise automatically A classifier for massive Chinese Web pages is implemented The experimental results demonstrate that the classifier is effective and can satisfy category navigation service of search engine