Bayesian-Graph-Convolutional-Network-for-Traffic-Prediction

TODO

• Code release
• Release the code for calculating the posteriori of the road network

Introduction

For the convenience of comparison, we integrate the proposed BGCN into an open library for urban spatial-temporal data mining, called LibCity.

For the implementation details of BGCN, please see the file "libcity/model/traffic_speed_prediction/BGCN.py".

Train and Test

First, download datasets following the instructions in LibCity.

Second, train and test the model using the following command:

python run_model.py --task traffic_state_pred --model BGCN --dataset PeMS07

The results are recorded in the folder "libcity/log".

Acknowledgement

This project is based on LibCity. Thanks for the awesome work.

Citation

Please cite the following paper if you use this repository in your reseach.

@article{FU2024127507,
title = {Bayesian graph convolutional network for traffic prediction},
journal = {Neurocomputing},
pages = {127507},
year = {2024},
issn = {0925-2312},
doi = {https://doi.org/10.1016/j.neucom.2024.127507},
url = {https://www.sciencedirect.com/science/article/pii/S0925231224002789},
author = {Jun Fu and Wei Zhou and Zhibo Chen},
keywords = {Traffic prediction, Bayesian, Generative model},
}

Contact

For any questions, feel free to contact: [email protected]