Abstract: User access patterns reflect the laws of the user browsing the sites and the interest of user's access. The common pattern discovery method is mining frequent traversal paths of the user from the user's access logs. As the mining method of the frequent traversal path based on Apriori algorithm needs frequent access to the database and produce large amounts of candidate items, this paper presents a new data structure of the traverse path tree to mine user's frequent traversal paths. Results show that the mining efficiency is improved comparing with the traditional method.