Efficient Incremental Mining of Qualified Web Traversal Patterns without Scanning Original Databases

Discovering web traversal patterns is an important issue in web usage mining with various applications like navigation prediction and improvement of website management. Since web data grows so rapidly and some web data may become out of date over time, we need not only consider the new data but also delete the old one to re-mine new web traversal patterns. To reduce the overhead of re-mining the web traversal patterns from the whole web data, an incremental mining approach is needed by using the previous mining results and computing new patterns just from the inserted or deleted part of the web data. In this paper, we propose an efficient incremental web traversal pattern mining algorithm named IncWTP_PLM (Incremental mining of Web Traversal Patterns by using Projected-database Link Matrix). Meanwhile, a special data structure named Projected-database Link Matrix is proposed to avoid scanning original database. Besides, the website structure is also considered in IncWTP_PLM such that each web traversal pattern discovered is qualified. The experimental results show that our algorithm outperforms other approaches substantially in terms of efficiency.