Traffic flow prediction based on temporal multigraph convolutional neural network

Traffic flow prediction plays an important role in intelligent transportation systems(ITS), but is challenged by the spatiotemporal complexity of traffic flow connections. In order to integrate the traffic temporal correlation, spatial correlation and semantic correlation in road network models, we propose a deep learning framework, Temporal Multi-Graph Convolutional Neural Network(T-MGCN) for traffic flow prediction. Firstly, we identify several semantic correlations and encode the non-Euclidean spatial correlations and heterogeneous semantic correlations between roads into multiple graphs. These correlations were modeled by a multi-graph convolutional neural network. Next, a recurrent neural network model is used to learn the dynamic characteristics of the traffic flow to obtain temporal correlations; Finally, a fully connected neural network is used to fuse spatio-temporal correlations and semantic correlations