Bayesian and Dempster-Shafer models for combining multiple sources of evidence in a fraud detection system

TL;DR

This paper compares Bayesian and Dempster-Shafer methods for combining evidence from multiple sources to improve fraud detection, highlighting their differences in handling uncertainty and conflict.

Contribution

It provides a detailed description of both methods and demonstrates their application in estimating a global score for fraud detection systems.

Findings

Dempster-Shafer handles uncertain and conflicting evidence effectively.

Bayesian requires prior and likelihood estimates, while Dempster-Shafer relies on posterior probabilities.

Both methods can be applied to improve evidence integration in fraud detection.

Abstract

Combining evidence from different sources can be achieved with Bayesian or Dempster-Shafer methods. The first requires an estimate of the priors and likelihoods while the second only needs an estimate of the posterior probabilities and enables reasoning with uncertain information due to imprecision of the sources and with the degree of conflict between them. This paper describes the two methods and how they can be applied to the estimation of a global score in the context of fraud detection.