Mining Web Usage Data for Automatic Site Personalization

Abstract

The ability to collect detailed usage data at the level of individual mouse clicks, provides Web-based companies with a tremendous opportunity for personalizing the Web experience of clients. Most current approaches to Web personalization include using static profile of users obtained through registration, and approaches based on collaborative filtering. These approaches suffer from the problems of the profile data being subjective, as well as getting out of date as the user preferences change over time. We present an approach to Web personalization based on Web usage mining, taking into account the full spectrum of data mining techniques and activities. We describe and compare Web usage mining techniques, based on transaction clustering and pageview clustering, to extract usage knowledge for the purpose of Web personalization. We also discuss how the extracted knowledge can be effectively combined with the current status of an ongoing Web activity to perform real-time personalization. This approach allows personalization to be achieved based on objective aggregate ``usage profiles{''} representing how users actually tend to use a site rather than based on subjective ratings or registration-based information.