Personalized Web Page Recommendation Using a Graph-Based Approach to Implicitly Find Influential Users

In this paper, we propose a novel graph-based approach for modeling the browsing data of Web users in order to understand their interests and their relationship with other users in the network. The aim was to identify users who are more influential while recommending pages to a network of users with similar interests. We call these users influential users and assign them an influence score that indicates the extent to which similar users follow their recommendations. By monitoring the browsing activity of influential users, we can identify their interest profiles as well as relevant pages quickly, and recommend these pages to users with similar interests. We call our proposed graph-based model a recommendation network. In this graph, nodes represent users and an edge between users u and v expresses the fact that u and v have similar interests, in particular the weight of the edge is the degree to which the user interest profiles match. Based on the graph, we build a recommendation system for Web pages, taking into account the influence of users in a network. Experimental results that measure the precision, with which recommended Web pages are visited by users, indicate that our system performs significantly better than traditional collaborative filtering-based recommender systems.