Utility-Based Web Path Traversal Pattern Mining

Web usage mining is to discover user traversal patterns of Web pages from Weblog records. Usually, a popular Website may register the Weblog records in the order of hundreds of megabytes every day, which provide rich information about the Web dynamics. Path traversal pattern mining discovers frequent sequential Web accessing patterns from Weblog databases. However, it fails to reflect the different impacts of different Web pages to different users. The difference between Web pages makes a strong impact on the decision-makings in Internet information service applications. Therefore, in this paper, we introduce "utility" into path traversal pattern mining problem. Utility is a measure of how "interesting" or "useful" a Web page is. As a result, it allows Web service providers to quantify the user preferences of different traversal paths. Two-Phase utility mining method is used to discover high utility path traversal patterns. We apply our proposed "high utility path traversal mining" algorithm on a real-world Weblog database, and compare the high utility path traversal patterns with the frequent traversal patterns by a traditional path traversal method. We demonstrated the interesting paths, as well as their significance to the decision making process.