Abstract: This paper mainly introduces a content-based web pages automatic classification system. Especially, it discusses the method of web pages key words extraction,the auto matic classification algorithm of Chinese web pages.The algo rithm decides on member ship function of indexing descriptors belonging in each of classification by statistics and classifies a rchives with fuzzy relation.The results of the exper iment show that the two impor tant factors,which are used for evaluating the automatic classification algo rithm,precision and recall,are improved.