TrafficTL

The code for paper "Traffic Prediction with Transfer Learning: A Mutual Information-based Approach"

Notes

1. This version is a reconstruct code for TrafficTL, so it omits some complicated operations in papers like replace nodes in source city with nodes in target city.

Requirements

include but not limited

torch
tqdm
dtaidistance
logging
pyyaml
numpy
sklearn
pathlib

Step 1

For data privacy, please apply follow line to generate sample data for going through whole process.

python data_generate.py

Step 2

if you want to go through the pipeline, please use next command.

python main.py --src_city 'src' --trg_city 'trg' --device 0

if you want to use it on your own data, please place your data on the file 'data' and specify data path in 'config.yaml'.

python main.py --src_city 'xxx' --trg_city 'xxx' --device 0

please use city name replace 'xxx'.

Thanks

Thanks for repository codes IIC and DCRNN

Citation

@ARTICLE{10105852,
  author={Huang, Yunjie and Song, Xiaozhuang and Zhu, Yuanshao and Zhang, Shiyao and Yu, James J. Q.},
  journal={IEEE Transactions on Intelligent Transportation Systems}, 
  title={Traffic Prediction With Transfer Learning: A Mutual Information-Based Approach}, 
  year={2023},
  volume={24},
  number={8},
  pages={8236-8252},
  doi={10.1109/TITS.2023.3266398}}