Abstract: In view of users commonly using tree browser pattern， this paper presents a new collection algorithm of user access path top-down to reduce the generation of a short path， and merges it into the users browser tree； based on the concept of cross-page access frequency， establishes the cross-page access frequency matrix for collaborative filtering recommendation system core processing data source. Taking the cross-page user access weight for the judgment basis， it achieves the recommendation of the page associated with the user browser.

Key words: access path, tree, recommended system, collaborative filtering