Decision-Focused Learning: When and Why Traditional Prediction Models Fail
For the operations research community, this tutorial clarifies the fundamental disconnect between prediction and decision quality, motivating a shift toward decision-focused methods.
This tutorial reviews decision-focused learning (DFL), showing that improved predictive accuracy does not guarantee better decision quality, and highlights why traditional statistical learning tools like uncertainty-driven data collection and Wasserstein distance are unsuitable for DFL, particularly in stochastic linear programming.
Plugging predictions of unknown parameters into downstream optimization problems, often referred to as the ``predict-then-optimize'' paradigm, has long been a standard approach in decision-making under uncertainty. However, improved predictive accuracy does not, in general, translate into improved decision quality. This disconnect has motivated growing interest in decision-focused learning (DFL) within the operations research community. This tutorial reviews recent developments in DFL and highlights key methodological insights, with a particular focus on stochastic linear programming as the downstream decision-making problem. We discuss why several widely used tools in traditional statistical learning are not directly suited to decision-focused settings and must be rethought, including (i) data collection strategies driven purely by predictive uncertainty and (ii) distributional distance measures such as the Wasserstein distance. We summarize properties of DFL that distinguish it from conventional predictive modeling and provide insights into the development of new decision-focused tools.