[2601.16632] Dual-Prototype Disentanglement: A Context-Aware Enhancement Framework for Time Series Forecasting

Computer Science > Machine Learning arXiv:2601.16632 (cs) [Submitted on 23 Jan 2026 (v1), last revised 30 Mar 2026 (this version, v3)] Title:Dual-Prototype Disentanglement: A Context-Aware Enhancement Framework for Time Series Forecasting Authors:Haonan Yang, Jianchao Tang, Zhuo Li View a PDF of the paper titled Dual-Prototype Disentanglement: A Context-Aware Enhancement Framework for Time Series Forecasting, by Haonan Yang and 2 other authors 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 outpu...