From Hybrid to End-to-End: Frameworks for Decision-Focused Financial Regime Modeling

Industrial Problems Seminar

Chenyu Yu
Quantic, Walleye Capital

Abstract

Financial markets move through distinct regimes, but regimes are hard to use as investment signals: they are usually labeled after the fact and loosely defined. This talk shows how to treat regimes as learnable, forecastable patterns that inform portfolio optimization, turning a macroeconomic concept into an investable signal. I will present two approaches: a hybrid framework that first identifies regimes by clustering and then forecasts them with a supervised model, and an end-to-end model that learns the regimes and the regime-specific predictive models jointly.