Li Chen, Melvyn Sim, Xun Zhang, Long Zhao, Minglong Zhou
We propose a new robust explainable prescriptive analytics framework that minimizes a risk-based objective function under distributional ambiguity by leveraging on the data collected on the past realizations of the uncertain parameters affecting the decision model and the side information that have some predictive power on those uncertainties.
Operations Research. https://pubsonline.informs.org/doi/abs/10.1287/opre.2023.0300
Zhiyuan Wang, Lun Ran, Minglong Zhou, Long He
We demonstrate that DRO and RS can share the same solution family under mild conditions. We establish the correspondence between the radius parameter in DRO and the target parameter in RS such that the optimal solutions to the two models are the same. We also extend the results to globalized-DRO and globalized-RS.
Manufacturing & Service Operations Management. https://pubsonline.informs.org/doi/abs/10.1287/msom.2023.0531
Chenyi Fu, Ning Zhu, Minglong Zhou
We develop a two-stage robust satisficing RSU location model featuring uncertain demands for computing tasks and random delays in task processing and transmission. The optimization model minimizes the riskiness of violating various capacity budgets and task latency thresholds. The numerical experiments illustrate that our robust model is computationally tractable and outperforms the deterministic model and queuing model in various metrics.
Production and Operations Management. https://journals.sagepub.com/doi/10.1177/10591478251339816
Melvyn Sim, Qinshen Tang, Minglong Zhou, Taozeng Zhu
We consider a general data-driven decision-making problem with covariate information. We build upon the robustness optimization framework recently proposed by Long et al. (2021), and we extend it to incorporate aspects of predictive analytics.
Operations Research. https://pubsonline.informs.org/doi/abs/10.1287/opre.2023.0199.
Zhuoyu Long, Melvyn Sim, Minglong Zhou
We present a framework for optimization under uncertainty called robustness optimization. The robustness optimization model seeks the most robust solution that achieves the target profit. Specifically, the measure of robustness is defined by the maximum level of model infeasibility that may occur relative to the magnitude of deviation of the realization of uncertainty from the nominal value. In other words, we allow infeasibility in achieving the target profit when the realization of uncertainty deviates from its nominal value, but we minimize the level of infeasibility whenever this occurs. We present tractable robustness optimization formulations for linear, combinatorial, adaptive linear, data-driven adaptive linear, and dynamic optimization models.
Operations Research. https://pubsonline.informs.org/doi/abs/10.1287/opre.2021.2238
Yu Wang, Yu Zhang, Minglong Zhou, Jiafu Tang
We study a robust surgery scheduling problem augmented by patient feature segmentation. We focus on the surgery-to-operating room allocations for elective patients and future emergencies. Using feature data, we classify patients into different types using machine learning methods and characterize the uncertain surgery duration via a feature-based cluster-wise ambiguity set. We propose a feature-driven adaptive robust optimization model that minimizes an overtime riskiness index, which helps mitigate both the magnitude and probability of working overtime.
Production and Operations Management. https://onlinelibrary.wiley.com/doi/10.1111/poms.13949
Minglong Zhou, Gar Goei Loke, Chaithanya Bandi, Glen Liau (NUHS), Wilson Wang (NUHS)
We model the intraday scheduling problem in orthopedic clinics where scheduled patients need to go through multiple consultations in different stations. We propose a new perspective to model the dynamics, which renders it possible to incorporate no-shows, walk-ins, patient re-entries, uncertain patient arrival times, and random transportation times among stations. To the best of our knowledge, such a realistic setting is not addressed in the existing literature. We then propose a two-stage stochastic optimization model for deriving optimal intraday scheduling.
Manufacturing & Service Operations Management. https://pubsonline.informs.org/doi/abs/10.1287/msom.2020.0959
Minglong Zhou, Melvyn Sim, Shao-Wei Lam (Singhealth)
We study the advance scheduling of ward admission requests in a public hospital, which affects the usage of critical resources such as operating theaters and hospital beds. We propose a new risk measure, the resource satisficing index (RSI), to characterize the risk of resource overutilization. We then propose a data-driven model that balances out these risks by minimizing the largest weighted RSIs, which, under our proposed partial adaptive scheduling policy, can be solved via a converging sequence of mixed-integer optimization problems.
Production and Operations Management. https://onlinelibrary.wiley.com/doi/abs/10.1111/poms.13799
Chenyi Fu, Minglong Zhou, Melvyn Sim, Kelvin Tan (Singapore MOH)
We consider a vaccination allocation problem under the lens of robustness optimization. We extend the classical epidemiological model and propose a tractable robustness optimization model, which aims to satisfy a cost constraint and to manage healthcare capacity under uncertainty.
Available at: http://www.optimization-online.org/DB_HTML/2021/06/8458.html
Qinshen Tang, Yu Zhang, Minglong Zhou
We develop a practical model to support repositioning decisions for (on-demand) platforms such as bicycle-sharing systems and free-float car sharing systems. From a risk mitigation perspective, we aim to satisfy a target service level as much as possible. We propose a practical model under a realistic setting including stochastic demand, travel destination duration, and repositioning duration.
Melvyn Sim, Long Zhao, Minglong Zhou
We investigate the connection between regularization, robust optimization, and robust satisficing. In simulation with more than twenty popular datasets, we show that our approach can be more competitive than LASSO.
Available at: ssrn.com/abstract=3981205
Minglong Zhou, Jussi Keppo, Esa Jokivuolle
We model the interaction among lawmakers, firms, and their employees under the risk of insider misconducts. We show how various parameters affect the firm’s incentives to monitor employees and the employees’ incentives to misconduct. Our model supports empirical evidence that middle management typically causes large-scale misconducts and also provides insights on how the lawmakers can design effective regulations.