A Graph Neural Network Social Recommendation Algorithm Integrating the Multi-Head Attention Mechanism

Collaborative filtering recommendation systems are facing the data sparsity problem associated with interaction data, and social recommendations introduce user information to alleviate this problem. Existing methods cannot express interest influence deeply, which limits performance of system. To address problem, in paper we propose a graph neural network algorithm integrating multi-head attention mechanism. First, based on user-item graph, is used learn high-order relationship between users items deeply extract latent features items. In process learning embedding vector representation mechanism introduced increase importance friends high influence. Then, make rating predictions for target according learned item vector. The experimental results Epinions dataset show that proposed method outperforms existing terms both Recall Normalized Discounted Cumulative Gain.