Research On Social Network Recommendation Algorithm Based On Trust Relationship

In recent years,with the rapid development of the Internet industry,human beings have gradually entered the era of big data.But at the same time,due to the explosive growth of the amount of information data,"information overload" has become one of the major concerns in the industry.Recommendation system as a solution to the problem of information overload has been widely used in the Internet.At present,some progress has been made in the research of recommendation algorithms.Many scholars have proposed recommendation algorithms that integrate user trust information,but there is still room for improvement in the recommendation performance of the algorithms.Through the combination of user trust and social characteristics,the recommendation performance will be improved.The main research work of this paper is as follows:(1)Aiming at the problem of data set sparsity in Social networks and the characteristics of different interests and preferences among users,this paper proposes a Recommendation System with Trust Cluster and User characteristic Social Regularization algorithm(RSTU algorithm for short).In the algorithm,the similarity between users and trusted users in the cluster is calculated by using the feature vector of the user cluster with trust relation to the same trusted user as the feature vector of the trusted user,so as to reduce the sparsity of trust relation between users.Aiming at the problem that the number of users in the user cluster is small and the characteristics of the trusted user cannot be well represented,the similarity between the user and the corresponding trusted user cluster is used to calculate the minimum distance to improve the interpretability of the user cluster.Then,according to the characteristics of users' different interests and preferences in different projects,the similarity between the potential characteristics of different users in the same scoring project is calculated,and new constraint parameters are given in combination with the social characteristics of users,so as to reduce the influence of users' interests and preferences on the recommendation effect.Finally,this algorithm was compared with three traditional algorithms PMF,So Reg and SVD++ by using Epinions,which is a social network data set.The experimental results show that RSTU algorithm performs best on the performance evaluation index RMSE compared with other algorithms,which proves that the performance of RSTU algorithm on the accuracy of scoring prediction is improved,and the sparsity of data sets is solved well,thus improving the recommendation effect.(2)Most recommendation methods based on social network make use of the trust relationship between users but ignore the influence of Similarity information between items on users' interests.This chapter proposes a Recommendation System with Trust Relation and Item Similarity(RSTI algorithm for short).In algorithm structure,the first use of users entropy and improve Jaccard similarity,the integrated similarity by users with common score projects between the local similarity and global similarity to constitute project integrated similarity,the user further integrated similarity and project integrated similarity,to recommend,its purpose is to solve the problem of data sparseness and the user cold start,improve the rating accuracy.Finally,Ciao was used to compare and verify the RSTI algorithm with three traditional recommendation algorithms PMF,So Rec and Trust SVD,and the experimental results show that RSTI algorithm performs better than other algorithms in the performance evaluation index RMSE on the score prediction and user cold start.