| Abstract: |
Web personalization can be described as any action that makes the Web
experience of a user customized to the user's taste, preferences, or information need. Existing approaches used by many Web-based companies,
as well as approaches based on collaborative filtering, rely heavily on explicit user input for determining the personalization actions. This
type of input is often a subjective description of the users by the users themselves, and thus prone to biases. Furthermore, the profile is
static, and its performance degrades over time as the profile ages. In this paper we describe an approach to automatic Web
personalization based on mining of usage data, taking into account a full spectrum of
data mining techniques and activities. A general architecture for usage-based Web personalization is presented, distinguishing between the
offline tasks of data preparation, transaction identification, and Web usage mining; and the online process of customizing Web pages based on a
user's active session. We describe several techniques for extracting aggregated usage knowledge, suitable for the purpose of Web
personalization, based on association rule discovery and clustering. We also propose and evaluate specific techniques for combining the
extracted knowledge with the current status of an ongoing Web activity to perform real-time personalization. |