Research On Collaborative Filtering Recommendation Algorithm Based On Dual Trust Networks

The popularization of Internet and the rapid development of electronic commercesystem bring the explosive growth of information, excessive amounts of information makeit is hard for users to find the products they need. The Recommendation System appearsunder such a background, the Collaborative Filtering is the most widely used technologyin Recommendation System. But due to the sparsity of user-item rating matrix, it isimpossible for the Collaborative Filtering to provide users with satisfactoryrecommendation results. It can alleviate the problems existing in the CollaborativeFiltering to some degree by introducing the trust relationships between users to the processof recommendation, but the user trust rating matrix also has the problem of sparsity, whichmakes it impossible to recommend accurately for users. Therefore, how to effectivelyimprove the quality of the Recommendation System by the trust information of users hasbecome a widely concerned issue. In this paper, on the basis of the research and analysisabout the present situation at home and abroad, and has a further research about theCollaborative Filtering algorithm based on trust network.Firstly, aiming at the sparsity problem of user trust rating matrix in the trust-basedCollaborative Filtering algorithm, a trust network expansion algorithm is proposed. Theuser similarities are calculated according to the trust ratings between users, and the set ofusers with high similarity is created for each user, and the degree of trust between differentusers is differentiated, then the new trust relationships are predicted by using thedifferentiated degree of trust and the sets of similar users, and the expansion of trustnetwork is realized.Secondly, aiming at the problem that the existing trust-based collaborative filteringalgorithms only consider the trust information comes from a single trust network, acollaborative filtering recommendation algorithm based on dual trust networks is proposed.The algorithm combining with trust propagation, according to the different locations ofusers in the dual trust networks, different strategies of trust calculation are used to get thetrust degree of target user to the co-rating user, then the trust degrees are treated as the weights of the prediction formula, and the ratings are predicted by using trust degrees takeplace of the similarities of traditional Collaborative Filtering recommendation algorithm.Finally, we prove the validity of the trust network expansion algorithm proposed inthis paper, and give the experimental evaluations and analysis of the collaborative filteringrecommendation algorithm based on dual trust networks compared with otherrecommendation algorithms.