Research On Graph Convolutional Network Recommendation Algorithm Integrating Social Relationships

With the advent of the era of big data,the amount of information has grown exponentially.As a tool for filtering information,the recommender system has become an indispensable part of people's daily life.It can filter out the information that people are interested in from a large amount of data.In recent years,with the rise of social networks,recommendation algorithms integrating social relations have attracted extensive attention of researchers.The social recommendation algorithm enriches the user's preference characteristics by combining the social relations between users,and recommends the set of items that their friends like for the user.In recent years,graph convolutional networks have attracted widespread attention due to their strong ability to learn graph representations.Social recommendation based on graph convolutional networks has become a research hotspot and achieved good results,but there are still some problems.On the one hand,most algorithms do not consider the heterogeneity of users in social graphs and user-item bipartite graphs,and lack deeper mining of graph structure information.On the other hand,in the case of sparse social graph and useritem interaction graph data,there is a lack of exploration of possible implicit links in the graph.In view of the above problems,the main work of this paper is as follows:Aiming at the heterogeneity of user nodes in user social graph and user-item bipartite graph,a high-order social collaborative filtering model based on graph convolutional network is proposed.The model uses social space aggregation and interaction space aggregation operations to collect high-order collaborative information between nodes from user social graph and user-item bipartite graph respectively,avoiding the influence of heterogeneous relationships on node feature representation learning.In order to achieve a deeper mining of the graph structure,this paper merges the user social graph and the user-item bipartite graph into a heterogeneous graph,and adds a convolution aggregation operation of the item nodes to the user's high-order social neighbor nodes.The information in the graph structure is fully utilized to further enrich the feature representation of item nodes.A data augmentation operation is proposed for the graph convolution operation.In the process of each layer of graph convolution,the possible implicit links in user-item bipartite graph and user social graph are dynamically added by using the similarity between nodes,so as to realize the mutual promotion of graph structure update and node embedding learning,which alleviates the data sparsity problem.