The 20th and 21st joint COTA International Conference of Transportation Professionals (CICTP 2020-21)

With the acceleration of urbanization in China, the concept of intelligent transportation is put forward, and traffic prediction plays an increasingly important role in intelligent transportation system. Timely and accurate traffic forecasting can make a better traffic management and alleviate traffic problems, such as traffic congestion, traffic pollution. In recent years, more and more scholars have devoted themselves to the study of traffic forecasting models, in order to improve the accuracy and effectiveness of forecasting. At the same time, graph data structure can well express the topology structure of traffic network, so graph model has more development space in the field of traffic prediction. The main purpose of this paper is to provide a comprehensive survey for the graph neural network in the field of traffic prediction. First of all, we divided the graph model framework into four categories, namely graph convolution networks, graph attention networks, graph auto-encoders and graph generative networks. Then, some related literatures are introduced around the four frames. Finally, some suggestions on the future development direction of the graph neural network are given.

CICTP