ISyE Seminar Series: Güzin Bayraksan

"A Residuals-Based Approach for Contextual Stochastic Optimization Under Decision-Dependent Uncertainty"

Güzin Bayraksan

• Professor and Associate Chair for Research in the Integrated Systems Engineering Department
• Affiliated faculty member of the Sustainability Institute and the Translational Data Analytics Institute
  • Ohio State University

About the Seminar:

Contextual Stochastic Optimization (CSO) has emerged as a way model and solve real-world decision-making problems by integrating Machine Learning (ML) with optimization under uncertainty. Many CSO models consider contexts (or side information) that are separate from the decisions. However, in many real-word applications, the decisions of the optimization model significantly affect the uncertain parameters, hence becoming contexts themselves. For example, pricing decisions impact the uncertain demand of a product.

In this talk, we first introduce a methodology for modeling such decision‑dependent uncertainty using a residuals-based stochastic optimization approach. We use ML to learn how decisions shape uncertainty and then build a distribution around these predictions using empirical residuals. We present a Sample Average Approximation (SAA) approach and its Distributionally Robust Optimization (DRO) variant. We investigate theoretical guarantees, including asymptotic optimality and finite sample results. We discuss an application in Electric Vehicle (EV) charging station location problem, where the customer demand depends on both external contexts (e.g., education level) as well as EV charging station location choices we make—for example, opening a charging station can increase the nearby demand. Through numerical experiments, we demonstrate the effectiveness of our proposed approach.

Related Papers:

Q. Zhu, X. Yu, and G. Bayraksan, “Residuals-based contextual distributionally robust optimization with decision-dependent uncertainty"

H. Sun, X. Yu, and G. Bayraksan, “Contextual Stochastic Optimization for Determining Electric Vehicle Charging Station Locations with Decision-Dependent Demand Learning”

About the Speaker:

Güzin Bayraksan is a Professor and Associate Chair for Research in the Integrated Systems Engineering Department and an affiliated faculty member of the Sustainability Institute and the Translational Data Analytics Institute at the Ohio State University. Her research interests are in optimization under uncertainty, particularly stochastic and distributionally robust optimization, using data-driven, contextual, and Monte Carlo simulation-based methods. She applies these models and methods to solve problems of critical societal interest in water, energy, and transportation systems. Her papers have appeared in top journals such as Mathematical Programming, SIAM Journal on Optimization and Operations Research, and her research has been supported by multiple grants from the National Science Foundation (NSF) and the Department of Energy (DoE). She is the recipient of INFORMS ENRE Best Publication Award in Environment and Sustainability, Lumley Research Award (OSU), NSF CAREER award, Five Star Faculty Award (UA), and the INFORMS Best Case Study award. She served as the Chair of Stochastic Programming Society (SPS) (2019-2023), the Vice Chair of Optimization under Uncertainty of INFORMS Optimization Society, and the President of the INFORMS Forum on Women in Operations Research and Management Science.

If you wish to be added to the ISyE Graduate Seminar Series emailing list, please email Event Coordinator Emily Rice at [email protected].