Skip to content

PyTorch Implementation of "Adaptive Multi-Scale Decomposition Framework for Time Series Forecasting"

License

Notifications You must be signed in to change notification settings

TROUBADOUR000/AMD

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Adaptive Multi-Scale Decomposition Framework for Time Series Forecasting

arXiv

Overview

The Adaptive Multi-Scale Decomposition Framework (AMD) is a cutting-edge solution for time series forecasting, incorporating three main components: the Multi-Scale Decomposable Mixing (MDM) Block, the Dual Dependency Interaction (DDI) Block, and the Adaptive Multi-predictor Synthesis (AMS) Block.

Prerequisites

To get started, ensure you are using Python 3.10. Install the necessary dependencies by running:

pip install -r requirements.txt

Data Preparation

Download the required datasets from Autoformer and iTransfomer. Organize the data in a folder named ./data as follows:

data
├── electricity.csv
├── exchange_rate
├── ETTh1.csv
├── ETTh2.csv
├── ETTm1.csv
├── ETTm2.csv
├── solar_AL.txt
├── traffic.csv
└── weather.csv

Training Example

All training scripts are located in the ./scripts directory. The details of the hyper-parameter settings are in Appendix C.4 in our paper. To train a model using the weather dataset, run the following command:

./scripts/Weather.sh

Citation

If you find this repository helpful, please cite our paper:

@article{hu2024adaptive,
  title={Adaptive Multi-Scale Decomposition Framework for Time Series Forecasting},
  author={Hu, Yifan and Liu, Peiyuan and Zhu, Peng and Cheng, Dawei and Dai, Tao},
  journal={arXiv preprint arXiv:2406.03751},
  year={2024}
}

Acknowledgements

We gratefully acknowledge the following GitHub repositories for their valuable contributions:

Contact

For any questions or inquiries, please submit an issue or contact us via email:

About

PyTorch Implementation of "Adaptive Multi-Scale Decomposition Framework for Time Series Forecasting"

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published