In order to improve the value of web browsing history data and evaluate peoples web browsing behavior, a method for browsing history data based on TF-IDF and Naive Bayes which is used to classify peoples behavior on the Internet is designed. During the experiment, the number of browsing history data is about one million, which is gotten from 945 students real web browsing behavior. About 84% students behavior can be classified eventually. This method gets good effect and improves the value of the web browsing history data and accuracy of crawling page information.