Continuous Belief Functions For Evidential Reasoning

Abstract

Some recently developed expert systems have used the Shafer-Dempster theory for reasoning from multiple bodies of evidence. Many expert-system applications require belief to be specified over arbitrary ranges of scalar variables, such as time, distance or sensor measurements. The utility of the existing Shafer- Dempster theory is limited by the lack of an effective approach for dealing with beliefs about continuous variables. This paper introduces a new representation of belief for continuous variables that provides both a conceptual framework and a computationally tractable implementation within the Shafer-Dempster theory.