Graph neural network for traffic forecasting: A survey

Traffic forecasting is important for the success of intelligent transportation systems. Deep learning models, including convolution neural networks and recurrent networks, have been extensively applied in traffic problems to model spatial temporal dependencies. In recent years, graph structures systems as well contextual information, introduced achieved state-of-the-art performance a series problems. this survey, we review rapidly growing body research using different e.g. convolutional attention various problems, road flow speed forecasting, passenger urban rail transit systems, demand ride-hailing platforms. We also present comprehensive list open data source codes each problem identify future directions. To best our knowledge, paper first survey that explores application created public GitHub repository where latest papers, data, will be updated.