Dual-Prototype Disentanglement: A Context-Aware Enhancement Framework for Time Series Forecasting

View PDF HTML (experimental)

Abstract:Time series forecasting has witnessed significant progress with deep learning. While prevailing approaches enhance forecasting performance by modifying architectures or introducing novel enhancement strategies, they often fail to dynamically disentangle and leverage the complex, intertwined temporal patterns inherent in time series, thus resulting in the learning of static, averaged representations that lack context-aware capabilities. To address this, we propose the Dual-Prototype Adaptive Disentanglement framework (DPAD), a model-agnostic auxiliary method that equips forecasting models with the ability of pattern disentanglement and context-aware adaptation. Specifically, we construct a Dynamic Dual-Prototype bank (DDP), comprising a common pattern bank with strong temporal priors to capture prevailing trend or seasonal patterns, and a rare pattern bank dynamically memorizing critical yet infrequent events, and then an Dual-Path Context-aware routing (DPC) mechanism is proposed to enhance outputs with selectively retrieved context-specific pattern representations from the DDP. Additionally, we introduce a Disentanglement-Guided Loss (DGLoss) to ensure that each prototype bank specializes in its designated role while maintaining comprehensive coverage. Comprehensive experiments demonstrate that DPAD consistently improves forecasting performance and reliability of state-of-the-art models across diverse real-world benchmarks.

Subjects:

Machine Learning (cs.LG); Artificial Intelligence (cs.AI)

Cite as: arXiv:2601.16632 [cs.LG]

(or arXiv:2601.16632v3 [cs.LG] for this version)

https://doi.org/10.48550/arXiv.2601.16632

arXiv-issued DOI via DataCite

Submission history

From: Haonan Yang [view email] [v1] Fri, 23 Jan 2026 10:33:34 UTC (550 KB) [v2] Tue, 27 Jan 2026 08:51:39 UTC (551 KB) [v3] Mon, 30 Mar 2026 08:13:17 UTC (586 KB)