Reserch On Trust Network Based Context Aware Recommendation Algorithm

In the field of recommender systems, collaborative filtering recommendation algorithm hasbeen widely studied, in addition to the traditional algorithm considering the binary relationshipbetween users and items, researchers also have proposed a series of improved methods to considercontextual information and trust information. However, trust-based collaborative filteringrecommendation algorithms take the trust information into account, ignoring the influence ofcontextual information on rating prediction. To the opposite, context-aware matrix factorizationapproaches as we know do not take trust information into consideration. The two types ofimproved methods ignore contextual information in which users and projects as well as trustinformation between users respectively, which may lead to deviate the user’s needs. Therefore, todeal with the question of how to consider both contextual information and trust informationsimultaneously, we further study the collaborative filtering algorithm in this paper.First of all, aiming at the problem that the two existing types of context-aware matrixfactorization methods do not consider the trust network between users, we improved the baselinepart and interaction part in the rating prediction formulas and propose two new rating predictionformulas.Secondly, aiming at the problems of implicit trust network construction in the existingtrust-based collaborative filtering method, using the technologies of trust network, trustpropagation and so on, we propose the user-based implicit trust network model and item-basedtrust network model from the aspect of user and item respectively.Finally, we integrate the above two improvements to get two trust network based context awarerecommendation algorithms and conduct comparative experiments with the existing two kinds ofrecommendation algorithms. Experimental results show that the two algorithms proposed in thispaper outperform the existing two kinds of recommendation algorithms in the predictionperformance.