Bridging Time and Frequency: A Joint Modeling Framework for Irregular Multivariate Time Series Forecasting

TL;DR

This paper introduces TFMixer, a novel joint time-frequency modeling framework for irregular multivariate time series forecasting, effectively capturing global periodic structures despite irregular sampling.

Contribution

The paper presents TFMixer, which integrates a learnable NUDFT and a query-based patch mechanism to improve forecasting of irregular time series.

Findings

Achieves state-of-the-art forecasting accuracy on real-world datasets.

Effectively captures seasonal patterns from irregular data.

Outperforms existing methods in handling non-uniform sampling.

Abstract

Irregular multivariate time series forecasting (IMTSF) is challenging due to non-uniform sampling and variable asynchronicity. These irregularities violate the equidistant assumptions of standard models, hindering local temporal modeling and rendering classical frequency-domain methods ineffective for capturing global periodic structures. To address this challenge, we propose TFMixer, a joint time-frequency modeling framework for IMTS forecasting. Specifically, TFMixer incorporates a Global Frequency Module that employs a learnable Non-Uniform Discrete Fourier Transform (NUDFT) to directly extract spectral representations from irregular timestamps. In parallel, the Local Time Module introduces a query-based patch mixing mechanism to adaptively aggregate informative temporal patches and alleviate information density imbalance. Finally, TFMixer fuses the time-domain and frequency-domain representations to generate forecasts and further leverages inverse NUDFT for explicit seasonal extrapolation. Extensive experiments on real-world datasets demonstrate the state--of-the-art performance of TFMixer.