Abstract: Traffic prediction is one of the essential issues of an intelligent transportation system. An accurate traffic prediction plays a vital role in traffic demands such as transportation operation adjustment, improvement of efficiency in logistics industry and public travel planning. In recent years, various deep learning frameworks for solving this problem have been proposed, and considerable prediction accuracy has been obtained. The traffic flow prediction model was summarized with graph convolutional network (GCN) in this paper, starting from the basic definition of graph convolution, covering frequency and spatial domain graph convolution, and introducing the basic principles of GCN. Subsequently, by introducing spatio-temporal graph neural networks, graph autoencoders, and graph attention networks, the development process of prediction models in this field was clarified, and the structure and characteristics of different prediction models were summarized. Based on the summary of commonly used datasets, the future research directions from the perspectives of application research, model research, and multi-source data fusion were discussed.