Using non-traditional approaches to statistical classification and regression in DSS model analysis

In a model-based decision support system, the decision maker initially has two primary tasks: finding the key parameters in the model and discovering how those key parameters, both individually and interactively, affect the solution. This paper presents an application of a non-traditional approach to statistical classification and regression to the inductive analysis of model output. Specifically, we describe the application of Ivakhnenko's group method of data handling to the identification of key model parameters and the discovery of a simplified polynomial metamodel, both of which frequently enhance the decision maker's understanding of the modeled environment.