A Survey: Deep Learning-based Time Series Forecasting

A official pytorch implementation for the paper: ' Deep Learning-based Time Series Forecasting' Xiaobao Song, Liwei Deng,Hao Wang*, Yaoan Zhang, Yuxin He and Wenming Cao (*Correspondence) PDF

🎯Introduction

📚Model Statics

❗️❗️❗️Tips: Due to our carelessness, we incorrectly reclassified the CNN-1D model to the Transformer-based category in the paper. We sincerely apologize for this mistake.

The following are the baseline models included in this project (continuously updated):

ARIMA PDF (IEEE Transactions on Power Systems 2003)
ETS PDF (Journal of statistical software 2008)
Autoencoder PDF (Neurocomputing 2014)
SAE PDF (IEEE Transactions on Intelligent Transportation Systems 2014)
CNN-1D PDF (IEEE Symposium Series on Computational Intelligence 2017)
TCN PDF Code (ArXiv 2018)
GRU PDF (Artificial Neural Networks and Machine Learning–ICANN 2018)
Nbeat PDF (Journal of biomedical informatics 2020)
LSTnet PDF Code (ACM SIGIR 2018)
LogTrans PDF (NIPS 2019)
DeepAR PDF (NIPS 2020)
AST PDF Code (NIPS 2020)
Reformer PDF (ICLR 2020)
SSDNet PDF (IEEE International Conference on Data Mining2021)
Informer PDF Code (AAAl 2021)
Autoformer PDF Code (NlPS 2021)
Aliformer PDF (ArXiv 2021)
TFT PDF Code (ArXiv 2021)
TDformer PDF Code (ArXiv 2022)
Dlinear PDF (AAAl 2022)
Fedformer PDF Code (lCML 2022)
NS-Transformer PDF Code (NIPS 2022)
Pyraformer PDF Code (ICLR 2022)
ETSformer PDF Code (ArXiv 2022)
PatchTST PDF Code (ICLR 2023)
Crossformer PDF Code (ICLR 2023)
Scaleformer PDF Code (ICLR 2024)
Triformer PDF (ArXiv 2024)
......

🧾Dataset Statics

Get Started

📝Install dependecies [Back to Top]

Install the required packages

pip install -r requirements.txt

👉Data Preparation[Back to Top]

We follow the same setting as previous work. The datasets for all the six benchmarks can be obtained from [Autoformer]. The datasets are placed in the datasets folder of our project. The tree structure of the files is as follows:

\datasets
├─electricity
│
├─ETT-small
│
├─exchange_rate
│
├─illness
│
└─traffic

🚀Run Experiment[Back to Top]

We have provided all the experimental scripts for the benchmarks in the ./scripts folder, which covers all the benchmarking experiments. To reproduce the results, you can run the following shell code.

 ./scripts/ETTh1.sh
 ./scripts/ETTh2.sh
 ./scripts/ETTm1.sh
 ./scripts/ETTm2.sh
 ./scripts/exchange.sh
 ./scripts/illness.sh
 ./scripts/traffic.sh

📧Contact

For any questions or feedback, feel free to contact Xiaobao Song or Liwei Deng.

🌟Citation

If you find this code useful in your research or applications, please kindly cite:

@article{song2024deep,
  title={Deep learning-based time series forecasting},
  author={Song, Xiaobao and Deng, Liwei and Wang, Hao and Zhang, Yaoan and He, Yuxin and Cao, Wenming},
  journal={Artificial Intelligence Review},
  volume={58},
  number={1},
  pages={23},
  year={2024},
  publisher={Springer}
}

Song, X., Deng, L., Wang, H. et al. Deep learning-based time series forecasting. Artif Intell Rev 58, 23 (2025). https://doi.org/10.1007/s10462-024-10989-8

🤝Acknowledgments

We express our gratitude to the following members for their contributions to the project, completed under the guidance of Professor Hao Wang:

Xiaobao Song， Liwei Deng，Yaoan Zhang，Junhao Tan，Hongbo Qiu，Xinhe Niu

TCCofWANG/Deep-Learning-based-Time-Series-Forecasting