A survey of recommendation system via Graph Neural Network and collaborative filtering

Recommendation systems play a vital role in today's information-rich digital environment, where users are overwhelmed with a vast number of choices and content. Because of the limitation of formal recommendation system models, Graph representation learning has become the mainstream. Among all different approaches, GNN-based CF is a state-of-art method that shows great performance. This study aims to offer a comprehensive overview of the fundamentals of GNN-based CF. Specifically, The study provides a taxonomy according to Graph Construction, Neighborhood Aggregation, Information Update, and Final Node Representation. Moreover, this survey summarizes different ways to enhancing. Meanwhile, a close classification of datasets are mentioned. This survey seeks to deepen the understanding of GNN-based CF, offering a roadmap for researchers and practitioners to navigate this exciting field and push the boundaries of recommendation system in our data-rich world.