A Model of Collaborative Filtering Based on Trust Communities in Complex Networks

In this paper, we propose a model of Social-based collaborative filtering that considers the reciprocal influence of users trust communites. This model consists mainly of three main steps: In Step 1, social communities in the trust social network are detected using a Genetic algorithm, and the items' similarity communities are computed. In Step 2, based on the depicted community structures and the rating data, a probabilistic model is developed to predict efficient recommendations. Finally, in Step 3, given a targeted user and a set of potential items, the learned probabilistic model is used to predict his/her preferential in order to propose him/her relevant recommendations. On the other hand, an experimental study was conducted to evaluate the cohesiveness of trust communities and the general accuracy of recommendations on Rich Epinion Dataset. The obtained results have shown that the proposed model outperforms some traditional methods of collaborative filtering.