CTMF: Context-Aware Trust-Based Matrix Factorization with Implicit Trust Network

Trust-aware recommender system can provide more accurate rating predictions than traditional recommender system by taking the trust relationships between users into consideration. Yet the state-of-the-art improved trust-aware collaborative filtering approach only considers the user-based implicit trust network model and the influence of trust information on rating prediction, ignoring the null value problem of local trust and the situation that people in different contextual conditions have different trust networks. The existing context-aware matrix factorization methods only consider the influence of contextual information on rating prediction, which are faced with the sparse initial rating matrix issue. To solve all the problems above, we propose two context-aware trust-based matrix factorization approaches to take both user-based implicit trust network model and item-based implicit trust network model into account and fully capture the influence of both context and trust information on rating. Experimental results on one real world dataset show that the two proposed approaches outperform the improved trust-aware approach and the existing context-aware matrix factorization methods in prediction performance.