基于双层网络结构综合考虑社交关系与偏好相似性的电影推荐方法

Abstract

Current movie recommendation methods lack consideration of multiple association relationships between users, resulting in sparse user friendship relationships and poor recommendation performance. To this end, this paper proposes a movie recommendation method that comprehensively considers social relationships and preference similarity based on a double-layer network structure. First, the social relationships and preference similarity relationships of movie website users are processed in layers according to different relationship types, and a social network is constructed through the follow relationships between users. Then, user preference similarity is measured based on the similarity of user-reviewed movie sets to construct a preference similarity network, thereby building a double-layer network structure that comprehensively considers user social relationships and preference similarity relationships. Finally, the Louvain community discovery method is used to perform dimensionality reduction analysis on large groups of users, and personalized recommendations are made for community users based on the influence of neighbor users. Experimental analysis on Douban data shows that the proposed method can effectively alleviate the sparsity problem of user social relationships, efficiently identify communities of interest to users, and provide target users with recommendation lists of higher novelty based on community neighbor evaluation information.