A Collaborative Filtering Recommendation Algorithm Based on Community Detection and Graph Neural Network

Recommendation system is an important module of many online systems. As one of the mainstream methods in the current recommendation system, the recommendation algorithm based on graph neural network can not only learn the cooperation signal between nodes, but also detect the nonlinear high-order information in node interaction. However, like the traditional collaborative filtering algorithm, the collaborative filtering method based on graph neural network only depends on the interaction information of existing users and projects, and the recommendation effect is poor in the scenario of sparse data. Although incorporating item attributes as auxiliary information into the recommendation algorithm can alleviate this problem to a certain extent, directly using the item attributes that users browse to construct user embedded vectors cannot contain users' implicit preferences. Therefore, this paper constructs a user-user network based on the interaction information between users and projects, analyzes it by using community detection, divides user groups with common interests, mines the potential attributes of users, and designs a collaborative filtering recommendation algorithm based on community detection and graph neural network. The research on two large real data sets shows that our method is superior to the standard recommendation method and the latest collaborative filtering recommendation method based on graph neural network.