Model recovers lost details for better long-term time forecasting
Recovering Lost Details: Multi-Scale Frequency Compensation for Long-Term Time Series Forecasting
Artificial Intelligence
Summary
Long-term forecasting models often lose important small details because they summarize data at different time scales. The authors present a new method called MWMixer that mixes information across multiple time scales to recover those lost details. This approach also adjusts how much each time scale contributes dynamically and ensures consistent predictions at every scale. Their experiments show that MWMixer performs well across many real-world datasets.
What this means in practice
- •For energy demand planners: Improve forecasting models to capture detailed fluctuations in electricity usage over many months for better resource management.
- •For financial analysts: Enhance stock or market predictions by recovering fine-scale temporal changes that traditional multi-scale models might miss.
Authors
Runmin Zou, Siyi Xie, Yaohui Huang, Yun Wang
Abstract
Long-term time series forecasting has made significant progress by leveraging multi-scale information to capture hierarchical temporal patterns and model long-range dependencies. However, temporal downsampling in existing multi-scale methods inevitably smooths detailed temporal fluctuations, and this information loss is further aggravated by their emphasis on dominant trends across scales, resulting in insufficiently expressive representations. To address this, we propose a Multi-Scale Wavelet Mixing (MWMixer) model, which incorporates a Bidirectional Frequency-Bands Mixing strategy to recover lost temporal details across scales, enabling complementary cross-scale information interactions. Then, a Dynamic Scale-Adaptive Fusion module learns time-varying weights for each scale to fuse multi-scale forecasts into the final prediction, enhancing the flexibility of multi-scale aggregation. In addition, a cross-scale consistency loss aligns each coarse-scale prediction with the interval-averaged fine-scale outputs, while a multi-scale supervision loss enforces prediction accuracy at each scale, promoting consistent learning across scales. Extensive experiments on seven real-world datasets demonstrate that MWMixer achieves competitive performance in long-term forecasting.