A Trust Network-based Collaborative Filtering Recommendation Strategy

As one of the most prevalent recommendation approaches, Collaborative Filtering (CF) provides users with personalized services according to similarity of their preferences, but seriously limited due to the Sparsity problem. A novel recommendation strategy is proposed in this paper to deal with this problem by introducing trust network in traditional CF process. The trust value is propagated through the trust network to match more neighbors for cold start users, and is combined with similarity to generate a compound weight to produce recommendations. Experiment results proved its efficiency obviously.