A Comprehensive Method for Model Credibility Measurement

Credibility is an integrated index to measure whether a simulation model is credible enough to a specific requirement. It is an overview of model accuracy, uncertainty, and performance throughout its whole life cycle following a particular simulation requirement. In this chapter, we present a framework based on a NASA standard to unify the complete process of model credibility measurement. As a significant amount of theories and methods have been proposed to implement VV and A for various models, we mainly focus on how to extract the credibility-relevant characteristics from the life cycle of a model and how to incrementally incorporate the existing methods and historical data to evaluate the credibility of a specific model during its real-time simulation. With the framework proposed in this chapter, both quantitative evaluation result and experts' scoring are combined to give a hierarchical view on model credibility. Moreover, the credibility evaluation process can be accelerated and made more intelligent using the incremental learning method.