Recommendation models for user accesses to web pages

Abstract

Predicting the next request of a user as she visits Web pages has gained importance as Web-based activity increases. There are a number of different approaches to prediction. Markov models and their variations, collaborative filtering models, or models based on pattern recognition techniques such as sequence mining, association rule mining, clustering user sessions or user, have been found well suited for this problem. In this paper we review these techniques and also highlight two new models that we have proposed. They consider the user access patterns to the pages as well as the time spent on these pages. We report experimental studies that show that the proposed methods can achieve a better accuracy than the other approaches. © Springer-Verlag Berlin Heidelberg 2003.