Analysis of simulation models with fuzzy graph based metamodeling

The analysis of complex simulation models using so-called metamodels offers reduced complexity and an understandable representation. In this paper we present an efficient algorithm that constructs a metamodel only from simulation data, so no a priori knowledge has to be included. Since no additional parameters have to be adjusted, the method is easy to use. It will be shown that the resulting system approximates the underlying model with an adjustable precision. In addition, the data can contain imprecise values or values with a corresponding confidence interval. This is especially well suited for simulation data due to its stochastic nature. The metamodel is represented in form of a Fuzzy Graph which allows the analyst to directly extract easy to interpret if-then rules. A real world token bus model was approximated with the proposed method. It is shown how the resulting Fuzzy Graph can be used to analyze this model and how the rule extraction leads to meaningful information about the model behavior.