Customized Sequential Designs for Random Simulation Experiments: Kriging Metamodeling and Bootstrapping

Jack Kleijnen
Department of Information Management
Tilburg University


Tuesday, November 30, 2004
4:15 - 5:30 PM
Terman Engineering Center, Room 498


Abstract:

In this talk I propose a novel method to select an experimental design for interpolation in random simulation. (Though I focus on Kriging, this method may also be applied to other types of metamodels, such as linear regression models.) Assuming that simulation requires much computer time, it is important to select a design with a small number of 'observations'-also called 'simulation runs'. The proposed method is therefore sequential. Its novelty is that it accounts for the specific input/output behavior-or response function-of the particular simulation at hand; i.e., the method is customized (application-driven). A tool for this customization is bootstrapping, which enables the estimation of the variances of Kriging predictions for inputs not yet simulated. The new heuristic is tested through the classic M/M/1 queueing and the (s, S) inventory simulation models. For these simulations the novel design indeed gives better results than Latin Hypercube Sampling (LHS) with a prefixed sample of the same size.

---

Operations Research Colloquia: http://or.stanford.edu/oras_seminars.html