Skip to content

Latest commit

 

History

History
49 lines (33 loc) · 1.79 KB

README.md

File metadata and controls

49 lines (33 loc) · 1.79 KB
Raw

Cross-variable Linear Integrated Enhanced Transformer for Multivariate Long-Term Time Series Forecasting (Client)

This is the official repo for Cross-variable Linear Integrated Enhanced Transformer for Multivariate Long-Term Time Series Forecasting (Client).

Getting Started

  1. Install Python >= 3.6, and install the dependencies by:
pip install -r requirements.txt

  1. You can obtain all the nine datasets from [Google Drive], [Tsinghua Cloud] or [Baidu Drive] provided in TimesNet and put them into the folder ./dataset.

  2. You can reproduce the experiment results through through the training scripts ./scripts/, and the name of our model's scripts is started with 'Client'.

# ETTh1
bash ./scripts/ETT_script/Client_ETTh1.sh
# ECL
bash ./scripts/ECL_script/Client.sh

  1. You can visualize the predictions of Client through the notebook 'visualization.ipynb'.

  2. The origin experimental results of mask series are shown in 'mask_result.csv', and the origin experimental results of LTSF are shown in 'result_of_Client.txt'.

Citation

If you find our repo useful, please cite our paper:

@misc{gao2023client,
      title={Client: Cross-variable Linear Integrated Enhanced Transformer for Multivariate Long-Term Time Series Forecasting}, 
      author={Jiaxin Gao and Wenbo Hu and Yuntian Chen},
      year={2023},
      eprint={2305.18838},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}

Acknowledgement

We appreciate the following repos for their valuable code base or datasets:

https://github.com/thuml/Time-Series-Library

https://github.com/cure-lab/LTSF-Linear