Graph neural networks in recommender systems

Abstract

As a way to alleviate the information overload problem arisen with the development of the internet, recommender systems receive a lot of attention from academia and industry. Due to its superiority in graph data, graph neural networks are widely adopted in recommender systems. This survey offers a comprehensive review of the latest research and innovative approaches in GNN-based recommender systems. This survey introduces a new taxonomy by the construction of GNN models and explores the challenges these models face. This paper also discusses new approaches, i.e., using social graphs and knowledge graphs as side information, and evaluates their strengths and limitations. Finally, this paper suggests some potential directions for future research in this field.