Impact of input data parameter uncertainty on simulation-based decision making

Simulation-based optimisation is popularly employed to determine the best system configuration of complex simulation models. However, it is assumed that the input distribution and parameters are known. In this paper, the input distribution parameter is assumed to be unknown and the impact of the same on the optimal system configurations obtained is investigated. Bootstrapping of the sample data is performed to generate multiple parameter estimates, each of which is used to obtain an optimal solution using regression based simulation-optimisation. It is proposed that the convex combination of the multiple optimal solutions got using different regression models gives a better estimate of the `true´ optimum. The above approach is illustrated using the classical (s, S) inventory policy.