Machine Intelligence and Threat Identification Systems

摘要

This report presents the results of a study of various evidential reasoningparadigms. The paradigms were evaluated for their efficiency and effectiveness in solving classification problems as they apply to electronic warfare threat identification. The paradigms tested were Bayesian networks, the Dempster-Shafer theory of evidential reasoning, and fuzzy evidential reasoning. Domain knowledge and evidence are discussed as they apply to evidential reasoning systems. The support generation techniques for each of the evidential reasoning paradigms is described. The mathematical computations needed to construct the measure of support are defined and illustrated. A Generic Classification Tool (GCT) is presented with an overview of the data structures, design, features, and architecture. The methodology employed and the results of the analysis of the reasoning paradigms is presented. The domain knowledge bases and evidence used in the analysis are described. The performance of the alternative support generation paradigms over a range of imprecise domain knowledge and information are compared. Keywords: Evidential reasoning, Classification problems, Expert systems, Artificial intelligence, Fuzzy reasoning. (Author) (kr)