Learning parameters in directed evidential networks with conditional belief functions

Abstract

Directed evidential networks with conditional belief functions are one of the most commonly used graphical models for analyzing complex systems and handling different types of uncertainty. A crucial step to benefit from the reasoning process in these models is to quantify them. So, we address, in this paper, the issue of estimating parameters in evidential networks from evidential databases, by applying the maximum likelihood estimation generalized to the evidence theory framework.