The mathematical theory of evidence — A short introduction

Abstract

Evidence or Dempster-Shafer theory is used to model information which is both uncertain and imprecise. Such a piece of information can be captured by the mathematical model of a hint. It is shown how hints can be combined and used to judge hypotheses by degrees of support and plausibility. Applications of this theory to statistical inference, diagnostics and risk analysis, and to decision analysis are discussed. The practical implementation of Dempster-Shafer theory depends on appropriate computational architectures both for modeling and for the inference mechanisms. A fundamental scheme for local computation in hypertrees is presented.