Last modified: 2019-09-02

#### Abstract

Simulation is used extensively for the modelling of large, complex, dynamic and stochastic systems, but tend to depend heavily on computer execution time. The severity of this problem has been lessened by the increasing execution speed of available computer hardware. However, the development of powerful and easy to use simulation languages has resulted in the increasing development of large and complex simulation models. Furthermore, real time applications of simulation models, requiring a quick response, are becoming more common. The inclusion of a simulation model as part of a bigger model might require numerous replications of the simulation model. This is, for example, true in the case of an optimization model using a simulation model in conjunction with a search algorithm. These kinds of models might still require extensive execution times and an applicable simulation meta-model might be useful.

This paper will discuss and illustrate the development of meta-models of a Monte Carlo simulation model using techniques such as multi-variable regression and neural networks.