To address the problem that graph convolutional network is easy to restrict the model's effective learning of space-time dependence in traffic flow prediction,this paper proposes a Space-Time Graph Con-volution Recurrent Network(ST-GCRN)model.Firstly,the time convolution layer is used to eliminate re-dundant time information.Secondly,the graph convolutional network is combined with the improved gated recurrent network to obtain the space-time dependence.Finally,the residual encoding and decoding struc-ture is added to solve the problem that the model training gradient disappears,thus improving the prediction accuracy and realize multi-step prediction.The experiment results on the California highway dataset show that the mean absolute error and the root mean square error of the model compared with the benchmark mod-el are reduced by 11％and 7.5％,respectively.
维普
万方数据