A New Agent-Oriented Approach for Personalized Web Management

Abstract

Finding techniques for constructing a personalized web is an important issue in the World Wide Web research. In this paper, we explore the problem of dynamical personalized web promotion-how to effectively interpret dynamic user trends and how to efficiently promote the most relevant pages to users. First, we present a simple idea based on dynamic transactions to effectively represent the changing trends of web users over time. Second, we formalize the dynamical personalized web promotion problem as a similarity matching between a dynamic transaction and a group of clustered web logs. The dynamic promoting mechanism is designed for this purpose where two algorithms, cluster identification and page identification, are also presented. Finally, we introduce our method for efficiently clustering web introduce our method for efficiently clustering web logs. Parts of the ideas above have been implemented, and the preliminary results are encouraging.