STSGAN: Spatial-Temporal Global Semantic Graph Attention Convolution Networks for Urban Flow Prediction

Accurate urban traffic flow prediction plays a vital role in Intelligent Transportation System (ITS). The complex long-term and long-range spatiotemporal correlations of pose significant challenge to the task. Most current research methods focus only on spatial local areas, ignoring global geographic contextual information. It is challenging capture information from distant nodes using shallow graph neural networks (GNNs) model correlations. To handle this problem, we design novel semantic graph-attentive convolutional network (STSGAN), which deep-level achieve simultaneous modelling First, propose (GACN) extract importance different features learn correlation regions temporal causal convolution structure (TCN) utilized relationships between long-short times, thus enabling an integrated consideration overall Several experiments are conducted two real-world datasets, results show that our approach outperforms several state-of-the-art baselines.