Beyond Static Forecasting: Toward Adaptive, Explainable, and Domain-Informed Time Series Models

Abstract: Real-world time series exhibit evolving dynamics, concept drift, and domain-specific structure that challenge static forecasting models. This talk presents a unified perspective on adaptive forecasting, spanning online explainable model selection, dynamic ensemble learning, and domain-informed neural forecasting. Through applications ranging from Industry 4.0 to astrophysics, I will discuss how adaptation, explainability, and scientific priors can be combined to move beyond static forecasting toward more robust and trustworthy time series models.