A Multi-Agent Recommender System Using Social Networks

This article presents a Multi-Agent approach for handling the problem of recommendation, considering the changing user's interests. Our objective is to propose an hybrid approach, based on a combination of collaborative and content-based filtering. We proceed in two steps. The first step consists on exploiting user information for social networks to update the user profile and to improve recommender system. Indeed, user profiles have an important role in multi-agent recommender systems. The information stored in them improves the system's generated recommendations. The second step consists on a decentralized peer-to-peer architecture. There is a natural link between peer-to-peer systems and multi-agent systems. Thus, as part of our approach, we view our distributed collaborative filtering system as a set of independent agents, an agent for each peer in the system. Each agent will then aggregate the information from several other agents to different degrees of reliability and realize the trust contextualization.