Using Combined Markov Chains to Predict WWW Navigation Paths

The Internet network provides access to several hundreds of million sites and expands at a rate of a million pages a day. Yet, given the rapid and anarchical nature of its evolution, Internet remains an information network lacking in both organization and structure, thus making effective information retrieval difficult. In the present paper, we are concerned with the automatic structural adaptation of web sites. We offer a tool for accessing hypermedia information that is contextualized and customized according to user profile. To this end, we make use of a particular form of Markov Models that we have called Combined Markov Chains. These specialized models have been developed for predicting web pages likely to be of interest to web users during their navigation. We also present some related applications of our model in areas such as the prediction of HTTP requests and the generation of virtual tours. KEYWORDS Collaborative Filtering, Prediction, Combined Markov Chains

• Type: Conference
• Year: 2003
• Where: IADIS
• Authors: Y. Hafri
• Keywords: Automatic Structural Adaptation, IADIS 2003, IADIS 2008, Markov Chains, Network Provides Access