Mining user access patterns based on Web logs

摘要

In this paper, different from usual order, not directly use the maximal forward reference path to mine sequence patterns but use DBSCAN algorithm to cluster the Web pages that have been accessed by users. Then, decide the Web page class that each page belongs to based on heuristic rules. Next, cluster the users who have the same interest in one or some kinds of Web pages. One user can belong to several classes, because the user may be interested in different types of Web pages. Finally, based on theory of sequence patterns mining, mine out user access patterns in each class by GSP algorithm. The benefit of using cluster methods is to find out layers' or classes' relationships from data even without any layer information of data. In this way, the user access patterns can be found more precisely