Mining User Traversal Patterns Based On Web Log

Web log mining is one of the most important research directions in the web mining research field. The aim of web log mining is to find out user traversal patterns of web sites. The process mainly includes four steps: data collection, log preprocess, pattern recognition and pattern analysis.We extract avail information from raw visit logs and insert them into databases. We ascertain each user based on heuristic rules, and dig out each user's visit sequence. Then cluster the pages that had been visited by users by DBSCAN algorithm. Next cluster the user with the same interest in one or some pages class. Finally, based on theory of sequence patterns mining, we mine out each class user's visit pattern by GSP algorithm.Through analysis of each class user's visit patterns, we can adjust the site's topology, improve the amount of visit and visit speed, build up adaptive sites, optimize strategy of page cache and prefetch, and monitor invalid login.