Open AccessWFTNet: Exploiting Global and Local Periodicity in Long-term Time Series ForecastingOpen Access
Author(s)
Peiyuan Liu,
Beiliang Wu,
Naiqi Li,
Tao Dai,
Fengmao Lei,
Jigang Bao,
Yong Jiang,
Shu-Tao Xia
Publication year2024
Recent CNN and Transformer-based models tried to utilize frequency andperiodicity information for long-term time series forecasting. However, mostexisting work is based on Fourier transform, which cannot capture fine-grainedand local frequency structure. In this paper, we propose a Wavelet-FourierTransform Network (WFTNet) for long-term time series forecasting. WFTNetutilizes both Fourier and wavelet transforms to extract comprehensivetemporal-frequency information from the signal, where Fourier transformcaptures the global periodic patterns and wavelet transform captures the localones. Furthermore, we introduce a Periodicity-Weighted Coefficient (PWC) toadaptively balance the importance of global and local frequency patterns.Extensive experiments on various time series datasets show that WFTNetconsistently outperforms other state-of-the-art baseline. Code is available athttps://github.com/Hank0626/WFTNet.
Language(s)English
Seeing content that should not be on Zendy? Contact us.
To access your conversation history and unlimited prompts, please
Prompt 0/10