Offline sequential learning via simulation

Abstract

Simulation has been widely used for static system designs, but it is rarely used in making online decisions, due to the time delay of executing simulation. We consider a system with stochastic binary outcomes that can be predicted via a logistic model depending on scenarios and decisions. The goal is to identify all feasible decisions conditioning on any online scenario. We propose to learn offline the relationship among scenarios, decisions, and binary outcomes. An Information Gradient (IG) policy is developed to sequentially allocate offline simulation budget. We show that the maximum likelihood estimator produced via the IG policy is consistent and asymptotically normal. Numerical results on synthetic data and a case study demonstrate the superior performance of the IG policy than benchmark policies. Moreover, we find that the IG policy tends to sample the location near boundaries of the design space, due to its higher Fisher information, and that the time complexity of the IG policy is linear to the number of design points and simulation budget.

Keywords:

• Offline sequential learning
• binary outcomes
• logistic model
• scenario-based feasibility determination

Acknowledgments

We thank the anonymous reviewers for the valuable comments.

Additional information

Funding

The work reported in this article was supported by the National Science Foundation of China under grant numbers 71971176 and 72101042, and the Fundamental Research Funds for the Central Universities under grant JBK2103010.

Notes on contributors

Haitao Liu

Haitao Liu received his BE and MS degrees in industrial engineering and management science from Sichuan University, China, in 2015 and 2018. Now, he is a PhD candidate in the Department of Industrial Systems Engineering and Management (ISEM) at the National University of Singapore (NUS). His research interests include simulation optimization and statistical learning.

Hui Xiao

Hui Xiao received a PhD degree in the Department of ISEM at NUS. Currently, he is a professor in the School of Statistics, Southwestern University of Finance and Economics, China. His research is devoted to simulation optimization, large-scale optimization and reliability modeling and optimization.

Haobin Li

Haobin Li received his PhD degree in the Department of ISEM at NUS. Currently, he is a senior lecturer in the Department of ISEM at NUS. His research interests are in simulation optimization and designing high performance optimization tools with application to logistics and maritime studies.

Loo Hay Lee

Loo Hay Lee received his PhD degree in engineering science from Harvard University, USA. He currently is a professor in the Department of ISEM at NUS. His research interests include logistics, vehicle routing, supply chain modeling, and simulation-based optimization.

Ek Peng Chew

Ek Peng Chew received his PhD degree in industrial engineering from Georgia Institute of Technology, USA. He currently is a professor in the Department of ISEM at NUS. His research areas are in port logistics and maritime transportation, simulation optimization and inventory management.