Spatio-temporal Graph Convolutional Neural Network: A Deep Learning Framework for Traffic Forecasting

This paper introduces Spatio-Temporal Graph Convolutional Networks (STGCN), a deep learning framework for the time-series prediction problem of traffic forecasting, which is crucial for urban traffic control and guidance. Because traffic flow is highly nonlinear and complex, and traditional methods cannot satisfy mid- and long-term prediction tasks while often neglecting spatial and temporal dependencies, the authors formulate the problem on graphs and build the model with complete convolutional structures instead of regular convolutional and recurrent units.

This fully convolutional design enables much faster training speed with fewer parameters. Experiments show that STGCN effectively captures comprehensive spatio-temporal correlations by modeling multi-scale traffic networks and consistently outperforms state-of-the-art baselines on various real-world traffic datasets, making it an influential approach for graph-based traffic and spatio-temporal forecasting.