Research on Graph Network Social Recommendation Algorithm Based on AGRU-GNN

With the widespread use of social network data, how to effectively utilize social relationships in recommendation systems has become a hot research topic. This paper proposes a graph network social recommendation algorithm based on the combination of Adaptive Gated Recurrent Unit (AGRU) and Graph Neural Network (GNN), called AGRU-GNN. The algorithm models users' historical behavior sequences through AGRU and models users' social relationships through GNN, achieving joint learning of social information and user behavior. Experimental results show that AGRU-GNN outperforms traditional recommendation algorithms in multiple real-world datasets, especially in scenarios with strong social influence. This research provides new insights for further optimization of social recommendation systems.