SWIFT: Mapping Sub-series with Wavelet Decomposition Improves Time Series Forecasting

SWIFT is a lightweight and efficient model for Long-term Time Series Forecasting (LTSF) that integrates wavelet decomposition, cross-band information fusion, and sub-series mapping techniques to improve forecasting accuracy, especially in resource-constrained environments.

---

🚀 Features

• Wavelet-based Lossless Downsampling: Uses Discrete Wavelet Transform (DWT) for effective time-frequency analysis.
• Cross-band Information Fusion: Learns a shared representation across different frequency bands.
• Minimalist Model Design: Employs a single-layer linear or shallow MLP for sub-series mapping.
• Superior Performance: Achieves state-of-the-art (SOTA) results with only 25% of the parameters of a standard linear model.
• Optimized for Edge Devices: Suitable for low-computation environments with fast inference speed.

---

🔧 Installation

Clone the repository and install dependencies:

git clone https://github.com/LancelotXWX/SWIFT.git
cd SWIFT
pip install -r requirements.txt

---

📝 Dataset

You can obtain the well pre-processed datasets from [Google Drive] or [Baidu Drive], in a file named 'all_datasets.zip'. Place the downloaded data in the folder./dataset.

---

🔥 Run

To run the experiment in paper:

bash scripts/SWIFT/etth1.sh
bash scripts/SWIFT/traffic.sh

---

📌 Citing SWIFT

If you find SWIFT useful in your research, please cite:

@article{xie2025swift,
  title={SWIFT: Mapping Sub-series with Wavelet Decomposition Improves Time Series Forecasting},
  author={Xie, Wenxuan and Cao, Fanpu},
  journal={arXiv preprint arXiv:2501.16178},
  year={2025}
}