Research On Similarity-based Movie Recommendation Algorithm

With the rapid development of technology and the proliferation of Internet users,the problem of information explosion has become more and more serious.For merchants,in the face of the skyrocketing volume of users and content,obtaining information on user preferences to achieve personalised recommendations reduces user churn.The volume of evaluation data is too small,which leads to data sparsity and cold start problems.In order to solve these problems,the introduction of auxiliary information is a hot topic of current research.By increasing the number of user-item interactions,attribute data between items,as well as social networks,user characteristics,sentiment analysis and contextual content are combined with the recommendation system.Among them,knowledge graphs,as graph structures,are rich in structural and semantic information,and combining graphs with auxiliary information can improve the performance of recommendation systems.Therefore,this paper combines knowledge graphs with Bayesian recommendation algorithms and graph attention networks to propose a similarity-based recommendation system,combining the latest research developments in current recommendation systems and improving and experimenting with the algorithms that currently exist.The research for this paper is as follows:1.In this paper,we propose a fusion recommendation model of knowledge graph and Bayesian recommendation algorithm.The knowledge graph can alleviate the data sparsity problem by adding the idea of ripples to the knowledge graph to obtain the set of user preferences by propagating over the knowledge graph.User preferences are obtained by Bayesian recommendation algorithm.The two algorithms are fused to obtain the TOP-K recommendation,and finally the algorithm is shown to have better recommendation performance through experiments with a comparative baseline.2.In order to improve the current problem of incomplete use of information in graphs,this paper proposes a graph attention recommendation algorithm based on knowledge graphs.In this algorithm,the user similar set-movie knowledge graph is used as the relationship information in the recommender system,and combined with the node and neighbour node aggregation information,different attention weight information is added to obtain a vector representation through aggregation,so that the higher-order information in the graph can be effectively utilized.Finally,the comparison with the baseline experimental evidence proves that the algorithm has better recommendation performance.