Abstract
Web usage mining analyzes web site traffic patterns in order to provide feedback on how they are being used. In this paper we present a new tool for web usage mining that is based on a linear ordering of the page transition matrix created from web server access logs. The ordering provides a measure allowing web pages to be categorized as origins, hubs or destinations according to their position in the ordering. It also provides a measure of the orderliness of web site traffic. This approach is applied to a university’s web site traffic over time and results are discussed. Comparing web site traffic immediately after a major change to the web site design and then two years later, the traffic is more ordered. Results from the linear ordering approach are also compared to a Markov steady-state analysis of the page transition matrix. The mathematical formulation of the problem and the pseudocode used for its solution is presented and its application for bandwidth or size-limited devices is presented.