A Valid Candidate Approach to Mining Bi-Directional Traversal Patterns on the WWW

Mining traversal patterns is one of important topics in Web mining. It focuses on how to find the Web page sequences which are frequently browsed by users. In this paper, we propose two algorithms for mining traversal patterns. For the first algorithm, SpeedTracer*-I, it is a revised version of the SpeedTracer algorithm. It directly generates and counts all candidate patterns from user sessions. Moreover, it improves the checking step when candidate patterns are generated. Next, based on the SpeedTracer*-I algorithm, we propose the SpeedTracer*-II algorithm, which improves the performance of the SpeedTracer*-I algorithm by decreasing the times to scan the database. From the simulation results, we show that the SpeedTracer*-I algorithm needs less processing time than the SpeedTracer algorithm. Moreover, the SpeedTracer*-II algorithm needs less processing time than SpeedTracer*-I and Apriori-like algorithms (e.g., FS and FDLP algorithms). (keywords: association rules, data mining, traversal patterns, Web mining, WWW)