Abstract
A simulation model is often used as a proxy for the real system of interest in a decision-making process. However, no simulation model is totally representative of the reality. The impact of the model inadequacy on the prediction of system performance should be carefully assessed. We propose a new metamodeling approach to simultaneously characterize both the simulation model and its model inadequacy. Our approach utilizes both simulation outputs and real data to predict system performance, and accounts for four types of uncertainty that arise from the unknown performance measure of the simulation model, simulation errors, unknown model inadequacy, and observation errors of the real system, respectively. Numerical results show that the new approach provides more accurate predictions in general.
Xiaowei Zhang
Associate Professor
My research research focuses on methodological advances in stochastic simulation and optimization, decision analytics, and reinforcement learning, with applications in service operations management, financial technology, and digital economy.