Predict-Then-Optimize

Follow this topic Ask about this topic

Predict-then-optimize studies pipelines that predict quantities such as costs, demand, or outcomes and feed those estimates into an optimization model to choose actions. Methods include decision-focused losses such as SPO+, differentiable and solver-free surrogates, bilevel or KKT reformulations, and robust or conformal approaches to uncertainty. They are used in settings including shortest paths, knapsack problems, portfolios, energy scheduling, and resource allocation. A central tension is that predictive accuracy may not reflect decision quality, while methods that improve decisions must also address computational scale, uncertainty, and fairness.

Latest papers

36 papers8 papers in the last four weeks, up 14% on the four weeks before. 0.0% of all new papers.

Globalized Adversarial Regret Optimization: Robust Decisions with Uncalibrated Predictions
Mar 26, 2026Jannis Kurtz, Bart P. G. van ParysRobust Optimization Predict-Then-Optimize
University of Amsterdam · CWI Amsterdam
Integrating Explainable Machine Learning and Mixed-Integer Optimization for Personalized Sleep Quality Intervention
Mar 14, 2026Mahfuz Ahmed Anik, Mohsin Mahmud Topu, Azmine Toushik Wasi +2Predict-Then-Optimize
Shahjalal University of Science and Technology, Sylhet, Bangladesh · Florida A&M University, Tallahassee, FL, United States · University of Oklahoma, Norman, OK, United States
LeJOT-AutoML: LLM-Driven Feature Engineering for Job Execution Time Prediction in Databricks Cost Optimization
Mar 9, 2026Lizhi Ma, Yi-Xiang Hu, Yihui Ren +2Automated Feature Engineering Automated Machine Learning
University of Science and Technology of China, Hefei, China · Lenovo, Beijing, China
Simulating Classification Models for Ex-Ante Evaluation of Predict-Then-Optimize Methods
Sep 2, 2025Pieter SmetPredict-Then-Optimize
KU Leuven, Department of Computer Science, NUMA · Gebroeders De Smetstraat 1, 9000 Gent, Belgium
Data-Driven Maintenance Optimization: Aligning Prognostic Learning with Decision Objective
Jun 24, 2025Zhuojun Xie, Adam Abdin, Yiping FangPredictive Maintenance Predict-Then-Optimize
Laboratoire Génie Industriel, CentraleSupélec, Université Paris-Saclay
Randomly Wrong Signals: Bayesian Auction Design with ML Predictions
Feb 12, 2025Ilan Lobel, Humberto Moreira, Omar MouchtakiGame Theory Predict-Then-Optimize
NYU Stern School of Business · FGV/EPGE Escola Brasileira de Economia e Finança

Mar 26, 2026J/K move · Enter open · S save

Globalized Adversarial Regret Optimization: Robust Decisions with Uncalibrated Predictions

Jannis Kurtz, Bart P. G. van Parys

University of Amsterdam · CWI Amsterdam

Robust Optimization Predict-Then-Optimize Minimax Regret

Open paper AskPDF arXiv

Optimization problems routinely depend on uncertain parameters that must be predicted before a decision is made. Classical robust and regret formulations are designed to handle erroneous predictions and can provide statistical error bounds in simple settings. However, when predictions lack rigorous error bounds (as is typical of modern machine learning methods) classical robust models often yield vacuous guarantees, while regret formulations can paradoxically produce decisions that are more optimistic than even a nominal solution. We introduce Globalized Adversarial Regret Optimization (GARO), a decision framework that controls adversarial regret, defined as the gap between the worst-case cost and the oracle robust cost, uniformly across all possible uncertainty set sizes. By design, GARO delivers absolute or relative performance guarantees against an oracle with full knowledge of the prediction error, without requiring any probabilistic calibration of the uncertainty set. We show that GARO equipped with a relative rate function generalizes the classical adaptation method of Lepski to downstream decision problems. We derive exact tractable reformulations for problems with affine worst-case cost functions and polyhedral norm uncertainty sets, and provide a discretization and a constraint-generation algorithm with convergence guarantees for general settings. Finally, experiments demonstrate that GARO yields solutions with a more favorable trade-off between worst-case and mean out-of-sample performance, as well as stronger global performance guarantees.

Latest papers

Globalized Adversarial Regret Optimization: Robust Decisions with Uncalibrated Predictions

Integrating Explainable Machine Learning and Mixed-Integer Optimization for Personalized Sleep Quality Intervention

LeJOT-AutoML: LLM-Driven Feature Engineering for Job Execution Time Prediction in Databricks Cost Optimization

Simulating Classification Models for Ex-Ante Evaluation of Predict-Then-Optimize Methods

Data-Driven Maintenance Optimization: Aligning Prognostic Learning with Decision Objective

Randomly Wrong Signals: Bayesian Auction Design with ML Predictions