Extracting rules from fuzzy simulation

Abstract

Abstract Expert knowledge about a domain is obtained through a complex array of informal statistics, heuristics, and pattern recognition. We propose that simulation modelling can play a central role in the construction of deep models for expert system knowledge. As part of our investigation, we have designed methods for fuzzy simulation which permit us to evaluate an expert's knowledge against mathematical models that are ultimately more detailed than causal methods or methods based on predicate logic. We use causal methods as stepping stones toward more accurate system models that evaluate the expert's knowledge. A sample grocery store queuing example is demonstrated, and a discrete event simulation model is obtained as a hypothetical deep model of the expert's knowledge. Our fuzzy simulation approach can be an aid to expert system researchers who are studying the acquisition of deep domain knowledge from an expert.