The control and transformation metric: a basis for measuring model complexity

Abstract

The purpose of this report is to develop a complexity metric suitable for discrete event simulation model representations. Current software metrics, based upon graphical analysis or static program characteristics, do not capture the influence on complexity stemming from the inherent dynamics of a model. A study of extant software metrics provides a basis for identifying desirable properties for model application. Various approaches are examined, and a set of characteristics for a model complexity metric is defined. A metric evolves from the recognition of the two types of complexity: transformation and control, both of which appear prominently in model representations. Experimental data are presented to verify that the Control and Transformation (CAT) metric reflects the desired behavior. Experimental evaluation supports the claim that the CAT metric is an improvement over existing software metrics for measuring the complexity of model representations.