TFDNet: Time–Frequency enhanced Decomposed Network for long-term time series forecasting

Abstract

Long-term time series forecasting is a vital task and applicable across diverse fields. Recent methods focus on capturing the underlying patterns from one single domain (e.g. the time domain or the frequency domain) without a holistic view to process long-term time series from the time–frequency domains. In this paper, we propose a Time-Frequency enhanced Decomposed Network (TFDNet) to capture both the long-term temporal variations and periodicity from the time–frequency domain. In TFDNet, we devise a multi-scale time–frequency enhanced encoder backbone with two separate trend and seasonal time–frequency blocks to capture the distinct patterns within the decomposed components in multi-resolutions. Diverse kernel learning strategies of the kernel operations in time–frequency blocks have been explored, by investigating and incorporating the potential different channel-wise correlation patterns of multivariate time series. Experimental evaluation of eight datasets demonstrated that TFDNet is superior to state-of-the-art approaches in both effectiveness and efficiency. The code is available at https://github.com/YuxiaoLuo0013/TFDNet.