The Validation Method of Simulation Model Based on K-Means Clustering and Fisher Discriminant Analysis

Usually, many simulation models of a system are provided. The most credible model should be selected. When the system only has a single output, some classic validation methods can solve the problem. But they become powerless when the system has multiple outputs with different data types. For solving the problem, the feature differences of each kind data were given, the simulation outputs were divided into k kinds of clusters based on K-means clustering, and which cluster the reference output belongs to was judged based on Fisher discriminant analysis. The simulation models whose outputs and reference output are in the same cluster are considered credible, and the model whose output is nearest to the reference output is the most credible one. In the application, the most credible model of the attitude control system of a missile was judged effectively by the method.