Simulation-Optimization under Uncertainty through Metamodeling and Bootstrapping

Most methods in simulation-optimization assume known environments, whereas this research accounts for uncertain environments combining Taguchi's world view with either regression or Kriging (Gaussian Process) metamodels (response surfaces). These metamodels are combined with Non-Linear Mathematical Programming (NLMP) to find a robust optimal solution. Varying the constraint values in the NLMP model gives an estimated Pareto frontier. To account for the variability of the estimated Pareto frontier, this research uses bootstrapping which gives confidence regions for the robust optimal solution. This methodology is illustrated through the Economic Order Quantity (EOQ) inventory-management model, accounting for the uncertainties in the demand rate and the cost coefficients