Web Recommendation using Semantic Web and Distributed Learning Automata and Graph Partitioning

Recommendation systems aim at directing users toward the resources that best meet their needs and interests. One of the challenging tasks in improving web recommendation algorithms is the simultaneous use of users’s activity log and hyperlink graph of the web site. In this paper, we propose a new recommendation algorithm based on semantic web and web usage data and hyperlink graph of a web site. In the proposed algorithm, a distributed learning automata learns similarity between web pages of a web site using web usage data and hyperlink graph of the web site and semantic web in URL address. Then, a usage based page rank for all pages of the web site is calculated using the probabilities of actions in the distributed learning automata. The proposed algorithm uses these information to build a Markov model which will be used to recommend new web pages for a user. Experiments show that the proposed method outperforms Association Rule Mining algorithm and the only learning automata based method reported in the literature in terms of precision and coverage.