550. A robust variant to the smart predict-then-optimize approach.
Invited abstract in session MC-12: Robust optimisation and its applications, stream Applications: AI, uncertainty management and sustainability.
Monday, 14:00-16:00Room: B100/8009
Authors (first author is the speaker)
| 1. | Belen Martin Barragan
|
| University of Edinburgh Business School, The University of Edinburgh | |
| 2. | Aakil Caunhye
|
| Business School, The University of Edinburgh | |
| 3. | Xuefei Lu
|
| Skema Business School |
Abstract
In this research, we develop and explore a modified version of the smart predict-then-optimize (SPO) strategy, which considers uncertainties in data prediction and inputs when optimizing. Building on the fundamental principles of the SPO model, our method focuses on refining predictions to reduce regret when those predictions shape the parameters of an optimization problem. We shift from a fixed, deterministic approach to one where data inaccuracies introduce uncertainty, and we apply robust optimization methods to address these uncertainties. Specifically, we study three types of robustness (worst-case robustness, strict robustness, and intermediate robustness) that tolerate varying levels of suboptimality and thus replicate different robustness-enforcing strategies. We assess our robust optimization models considering both uncertainties in the predictions and in the covariates. Our numerical results show significant out-of-sample performance improvements under randomly generated covariate disturbances, compared to the classic SPO approach, even when a small sample size is used.
Keywords
- Optimization under uncertainty
- Multi-level optimization
- Optimization for learning and data analysis
Status: accepted
Back to the list of papers