Machine learning and sinulation model specification

Abstract

In this paper we investigate the possibility of incorporating machine learning techniques in simulation modelling. The paper discusses the simulation process and provides an overview of machine learning relevant to this research. Adopting a strategy of learning from examples, a prototype system helps the analyst to specify the model of a real world system under study. The prototype system consists of a knowledge base, a learner module, a performance module, and an interface. The analyst gives examples of the system behaviour through the interface. The performance module uses a knowledge base to induce the system concepts. If the output is not acceptable, changes are made and the learner module uses these changes to upgrade and augment the rules in the knowledge base. This process continues until a satisfactory output is obtained.