Research Of Social Recommendation Based On User's Multi-faceted Trust Influence

With the development of the mobile internet and information technology, people have entered into the info-overloaded era from the era which lacks information. Facing the huge amount of information, it is impossible for people to find those useful contents accurately and quickly. As an effective tool to implement information filtering, personalized recommender systems solve the problem of information overload to some degree, while it still has cold start and data sparse problems. As a socially oriented internet service platform which connects the world, social network amazes the field of recommendation. More and more people start to exploit trust relationships in social data to do social recommendation. Social recommendation alleviates cold start problem in recommendation to some extent, while it usually only involves the local influence between neighbors, and always assumes single and homogeneous trust relationships between users, ignoring user's global influence and multi-faceted trust relationships.Focusing on the problems of data sparse and single formula mode to compute similarity, combining with the status quo of the research at home and abroad, taking full account of local and global influence, and exploiting multi-faceted trust relationships to implement studies on social recommendation, this paper proposes a social recommendation algorithm based on user's multi-faceted trust influence, which improves the calculation method of transition probability in random walk model. Experiments demonstrate the effectiveness of the proposed method. Then on this basis, the paper designs a prototype system as well.This paper mainly includes the following work:(1) State the research status of recommender systems, and analyze the main problems faced by social recommendation in combination with the correlative technologies.(2) For the problem of ignoring user's global influence in social recommendation, this paper takes full account of user's local and global influence, and corrects the computational method of user's influence. Based on that, we propose a social recommendation algorithm based on user influence walk model.(3) For the problem of ignoring multi-faceted trust relationships in social recommendation, when comes to further analysis of user's local influence, multi-faceted trust relationships are considered. Therefore, this paper proposes the concept of user's multi-faceted influence, and corrects the computational method of user's local influence. As a result, the paper proposes the social recommendation algorithm based on user's multi-faceted trust influence, which not only involves user's local and global influence, but also the multi-faceted trust relationships.(4) According to the research of social recommendation based on user's multi-faceted trust influence, this paper designs a prototype system, and applies it into a comprehensive online review platform.