User-centric Trust-based Recommendation

Recommender Systems are widely used to achieve a pre-selection of items among a constantly growing variety of items. Last generation of recommender systems take into account trust between users. In this article, we propose several trust-based recommendation formulas that keep centered around the end-user and thus restrict information sharing to the user vicinity. Our proposed RS does not need any global knowledge: it limits data exchange to trusted friendship relations. A comparison of the proposed recommendation formulas with more classical ones is finally proposed, based on two kinds of simulation.