Recommendation Algorithm Based On Trust Network And Complete Tripartite Graph

With the popularity of computers and the rapid development of the Internet,the amount of information is explosive growth,and it is more and more difficult for users to find the information they need from the vast amount of data,so the personalized recommendation technology came into being,and the collaborative filtering algorithm is the most widely used in the industrial sector.Personalized recommendation based on graph is a collaborative filtering technology,and now introduces social tags,personalized recommendation based on the graph theory also developed from the original bipartite graph to the user-item-tag tripartite graph.At the same time,with the emergence of social network,the concept of trust has been introduced into the recommendation system.Undoubtedly,recommendation system based on user trust network makes the results more credible and can be explained.Collaborative filtering algorithms still have many problems in the field.In this paper,we conducted a research on trust network based on the user to fill the scoring matrix to solve some questions such as data sparse,cold-start and trust problem;and this paper proposed a complete tripartite graph model combined by material diffusion and heat transfer,aiming to solve the problem that the accuracy and diversity can not be compatible each other.And we recommended users,items and tags together finally.Main works in this paper is listed as follow:Firstly,this paper proposed a filled score matrix calculated by user trust matrix and similarity matrix based on trust network,aiming to solve the current situation,such as data sparse,cold-start and low trust degree in collaborative filtering recommendation field.Secondly,the relationships among users,items,labels were considered insufficiently in tripartite graph and the accuracy and diversity can not be compatible each other,so this paper presented complete tripartite graph model combined material diffusion algorithm and heat propagation algorithm,it considers both the accuracy and diversity.Followed,this paper combined the trust network based user in conjunction with complete tripartite graph.In order to improve the recommendation results' interpretability and trustworthiness,finally put forward a joint recommendation mechanism with items,users and tags.Finally,this paper selected MoiveLens dataset as the experimental data sources,compared the recommendation effects of traditional collaborative filtering recommendation to our recommendation model in the paper,so that it can prove the achievement that we get.