songjiang0909/deepfs

Code for paper "Bridging Self-Attention and Time Series Decomposition for Periodic Forecasting"

DeepFS

This is the implementation of our paper "Bridging Self-Attention and Time Series Decomposition for Periodic Forecasting", published at CIKM'22.

Data

The original data are from this repo, credits and copyrights belong to the authors!

How to run?

• Step1 (run):
  • cd ./src
  • python main.py --arguement arguement_values
  • See explanations for other arguements and parameters in main.py.

The prediction and trained models are stored under the result folder.

Contact

Song Jiang [email protected]

Repo reference

Informer

Bibtex

@inproceedings{deepfs2022,
  title={Bridging Self-Attention and Time Series Decomposition for Periodic Forecasting},
  author={Song Jiang, Tahin Syed, Xuan Zhu, Joshua Levy, Boris Aronchik, Yizhou Sun},
  booktitle={Proceedings of the 31st ACM International Conference on Information & Knowledge Management},
  year={2022}
}

deepfs's Issues

How to plot the periodicity-weight figure in the paper?

Thanks for sharing. I have some questions about how to plot the periodicity-weight figure in the paper?
Do you use weights(weights_pred) of the output? Or something else?
If it is weights, how to deal with it to plot the figure in the paper? after all, there are negative values in weights.
At last, can you provide the code to draw this figure if possible?