A Comprehensive Investigation on Graph Neural Networks: Models, Applications, and Future Challenges

Graphs, as a kind of complex data structure, are getting wide attention these days, requiring more research on relative models to extract information and make predictions based on graphs. Especially, Graph Neural Network (GNN) shows high performance in many areas. This paper surveys both traditional and advanced GNN models, with their core principles and creative features. It also introduces real-life applications of GNN models, including social recommender systems, drug discovery and traffic prediction. Additionally, this paper discusses current bottlenecks of GNN and prospects of possible improvements. By reviewing GNN from different aspects, this paper tries to provide an entire perspective on this direction for readers. People now encounter many more types of graphs than before, ranging from urban road networks to social platforms, and even molecular structures in biology. By connecting these scenarios, this paper emphasizes the growing importance of GNN research and its potential to transform both scientific discovery and industrial applications.