A taxonomy of model abstraction techniques

Model abstraction is a method for reducing the complexity of a simulation model while maintaining the validity of the simulation results with respect to the question that the simulation is being used to address. Model developers have traditionally used a number of abstraction techniques, and simulation researchers have conducted formal research to build a theoretical foundation for model manipulation. More recently, researchers in the artificial intelligence (AI) subfield of qualitative simulation have also been developing techniques for simplifying models, determining whether models results are valid, and developing tools for automatic model selection and manipulation. Metamodeling can also be considered as an abstraction technique. The purpose of the paper is to provide a taxonomy of abstraction techniques drawn from these fields. This taxonomy provides a framework for comparing and contrasting various abstraction techniques.