Approaches to management of uncertainty in an expert system

Abstract

The problem of reasoning with uncertainty in expert systems has become an issue of increasing interest. In this article, the development of EMERGE, an expert system for chest pain analysis, is traced. Modifications to the system to utilize new techniques are analyzed. In particular, the incorporation of techniques from approximate reasoning is discussed as applied in the chest pain rule base to take into account degrees of presence of symptoms and unequal contribution of antecedents. Specifically, theoretical results in approximate reasoning by Yager are applied in the EMERGE expert system in an attempt to improve its performance in borderline cases.