TimeEmb: A Lightweight Static-Dynamic Disentanglement Framework for Time Series Forecasting

2025-09-30

AI Summary

Time series forecasting faces challenges from temporal non-stationarity, where existing methods struggle to disentangle time-invariant structural patterns from dynamically evolving components, leading to performance degradation under distributional shifts. To address this, we propose TimeEmb—a lightweight static-dynamic disentanglement framework. It employs a global embedding module to capture long-term, stationary representations, and introduces a frequency-domain filtering mechanism inspired by spectral analysis to isolate and model multi-scale time-varying dynamics in the frequency domain. TimeEmb is plug-and-play, requiring no additional labels or architectural modifications to base models. Evaluated across multiple real-world benchmarks, it consistently outperforms state-of-the-art methods while reducing computational overhead. These results validate the effectiveness, robustness, and generalizability of the static-dynamic disentanglement paradigm for time series forecasting.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Planning, Routing, and Scheduling: Temporal Planning
• Intelligent Robots: Embodied AI

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Temporal non-stationarity, the phenomenon that time series distributions change over time, poses fundamental challenges to reliable time series forecasting. Intuitively, the complex time series can be decomposed into two factors, ie time-invariant and time-varying components, which indicate static and dynamic patterns, respectively. Nonetheless, existing methods often conflate the time-varying and time-invariant components, and jointly learn the combined long-term patterns and short-term fluctuations, leading to suboptimal performance facing distribution shifts. To address this issue, we initiatively propose a lightweight static-dynamic decomposition framework, TimeEmb, for time series forecasting. TimeEmb innovatively separates time series into two complementary components: (1) time-invariant component, captured by a novel global embedding module that learns persistent representations across time series, and (2) time-varying component, processed by an efficient frequency-domain filtering mechanism inspired by full-spectrum analysis in signal processing. Experiments on real-world datasets demonstrate that TimeEmb outperforms state-of-the-art baselines and requires fewer computational resources. We conduct comprehensive quantitative and qualitative analyses to verify the efficacy of static-dynamic disentanglement. This lightweight framework can also improve existing time-series forecasting methods with simple integration. To ease reproducibility, the code is available at https://github.com/showmeon/TimeEmb.

Problem

• Disentangling static and dynamic components in time series
• Addressing temporal non-stationarity in distribution shifts
• Improving forecasting accuracy with lightweight computational framework

Innovation

• Decomposes time series into static and dynamic components
• Uses global embedding for time-invariant representations
• Applies frequency-domain filtering for time-varying patterns