Skip to content

Attention Based Spatial-Temporal Graph Convolutional Networks for Traffic Flow Forecasting, AAAI 2019, pytorch version

Notifications You must be signed in to change notification settings

wanhuaiyu/ASTGCN-r-pytorch

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ASTGCN

Attention Based Spatial-Temporal Graph Convolutional Networks for Traffic Flow Forecasting (ASTGCN)

image-20200103164326338

This is a Pytorch implementation of ASTGCN and MSTCGN. The pytorch version of ASTGCN released here only consists of the recent component, since the other two components have the same network architecture.

Reference

@inproceedings{guo2019attention,
  title={Attention based spatial-temporal graph convolutional networks for traffic flow forecasting},
  author={Guo, Shengnan and Lin, Youfang and Feng, Ning and Song, Chao and Wan, Huaiyu},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={33},
  pages={922--929},
  year={2019}
}

Configuration

Step 1: The loss function and metrics can be set in the configuration file in ./configurations

Step 2: The last three lines of the configuration file are as follows:

loss_function = masked_mae
metric_method = mask
missing_value = 0.0

loss_function can choose 'masked_mae', 'masked_mse', 'mae', 'mse'. The loss function with a mask does not consider missing values.

metric_method can choose 'mask', 'unmask'. The metric with a mask does not evaluate missing values.

The missing_value is the missing identification, whose default value is 0.0

Datasets

Step 1: Download PEMS04 and PEMS08 datasets provided by ASTGNN.

Step 2: Process dataset

  • on PEMS04 dataset

    python prepareData.py --config configurations/PEMS04_astgcn.conf
  • on PEMS08 dataset

    python prepareData.py --config configurations/PEMS08_astgcn.conf

Train and Test

  • on PEMS04 dataset

    python train_ASTGCN_r.py --config configurations/PEMS04_astgcn.conf
  • on PEMS08 dataset

    python train_ASTGCN_r.py --config configurations/PEMS08_astgcn.conf

About

Attention Based Spatial-Temporal Graph Convolutional Networks for Traffic Flow Forecasting, AAAI 2019, pytorch version

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%