Research On Recommendation Algorithms Based On Trust Mechanism And Social Network

With the rapid development of information technology,mankind has gradually entered an era of interconnection.The amount of data of all kinds of information has increased exponentially.Traditional recommendation algorithms have more and more difficultly in dealing with the problem of "information overload".In the research of recommendation algorithms,researchers have proposed various hybrid recommendation algorithms,hoping to improve the recommendation performance by using more additional information.Among them,the research of recommendation algorithm which integrates user trust information has made great progress.SoReg,TCRec and other representative algorithms have been proposed.By combining trust information,the recommendation performance of TCRec and other algorithms has been improved,but the recommendation performance of the algorithm still has a long way to go.The reasons are as follows.(1)The social information provided in the datasets is so sparse that the trust factors used in TCRec and other algorithms don't contribute much to the improvement of the recommendation performance of the algorithm;(2)The algorithms ignore similarity of users' interests,and only choose user cluster members based on social information in datasets,which results in poor recommendation effect.(3)The algorithms only use the social relationship in and out degree to define the trust relationship,which can't truly reflect the relationship between users.Focusing on the shortcomings of TCRec algorithm,this paper proposes two improved TCRec algorithms:I-TCRec algorithm and TSCRec algorithm,and evaluates the algorithm based on public datasets.The main work and contributions of this paper are as follows:(1)In this paper,TCRec algorithm is improved and I-TCRec algorithm is proposed.In order to consider user trust in non-social information datasets,the algorithm uses a comprehensive user trust calculation method which combines user entropy difference and implicit user trust.Considering the common rating items among users,the user cluster members can be selected more accurately and the situation of high trust and low interest similarity can be reduced.The algorithm also proposes a global user trust calculation method to better describe the characteristics of trust clusters.Experiments show that the performance of the I-TCRec algorithm is better than that of the traditional recommendation algorithm in ratng prediction accuracy,which proves the performance of the algorithm.(2)In this paper,another improved algorithm TSCRec is proposed.The algorithm defines another method of computing user trust based on user social information provided in datasets and comprehensive user trust in I-TCRec algorithm.This algorithm greatly reduces the sparsity of trust information.In order to choose user cluster members,the algorithm uses user nodes and item nodes to construct a bipartite graph,and defines the weights between nodes by using user-to-user relationship and user-to-item relationship.Through the improved PersonalRank random walk algorithm,the user cluster members of target users are defined.Experiments show that the performance of TSCRec algorithm is better than that of several reference recommendation algorithms based on social networks,which proves the performance of TSCRec algorithm.