STGCN-pytorch

This is a Pytorch implementation of Spatio-Temporal Graph Convolutional Networks(STGCN) model from the paper "Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecasting"

Detailed analysis：STGCN

Model architecture

Requirements:

• python 3.6
• pytorch 1.3.1

Usage

$ python main.py

Code Organization

• code
  • models
  • utils
  • metr_data
  • main.py

Reference

Yu B, Yin H, Zhu Z. Spatio-temporal graph convolutional networks: A deep learning framework for traffic forecasting[J]. arXiv preprint arXiv:1709.04875, 2017.

Other implementations

[1] TensorFlow：VeritasYin/STGCN_IJCAI-18

[2] mxnet：Davidham3/STGCN