nitheshj/DynGWN

Dynamic Graph WaveNet

Dyn-GWN: Time-Series Forecasting using Time-varying Graphs with Applications to Finance and Traffic Prediction

This is the offical repo of the ICAIF 2023 paper Dyn-GWN: Time-Series Forecasting using Time-varying Graphs with Applications to Finance and Traffic Prediction

Dyn-GWN is an adaptation of GWN for dynamic graphs.

Requirements

This code has been tested with PyTorch 2.0.

Structure of the repo

Scripts to run the algorithms are located in scripts/. The current code supports the following datasets: Financial: Stock Volatility, Traffic: METR-LA, PEMS-BAY.

Citing DynGWN

If you find DynGWN useful in your research, please consider citing the following paper.

@inproceedings{Ibrahim2023,
    author = {Ibrahim, Shibal and Tell, Max and Mazumder, Rahul},
    title = {Dyn-GWN: Time-Series Forecasting using Time-varying Graphs with Applications to Finance and Traffic Prediction},
    year = {2023},
    publisher = {Association for Computing Machinery},
    address = {New York, NY, USA},
    booktitle = {4th ACM International Conference on AI in Finance},
    numpages = {9},
    keywords = {spatio-temporal modeling, time-series forecasting, time-varying graphs, graph neural networks},
    location = {New York, NY, USA},
    series = {ICAIF '23}
}