I tried to construct the Pytorch-version implementation of the spatio-temporal graph convolutional network proposed in Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecasting by Bing Yu, Haoteng Yin, Zhanxing Zhu.
The author's original implementations in TensorFlow can be found here and my code structure is borrowed from him (for details about the prediction problem, model and dataset, please refer to the link below):
- PyTorch
- Pandas
- Scipy
- Numpy
- Matplotlib
The dataset of 228 routes mentioned in the author's documents are included in the 'dataset' directory.