Global sensitivity analysis of uncertain parameters in Bayesian networks

Traditionally, the sensitivity analysis of a Bayesian network studies the impact of individually modifying the entries of its conditional probability tables in a one-at-a-time (OAT) fashion. However, this approach fails to give a comprehensive account of each inputs' relevance, since simultaneo...

Full description

Bibliographic Details
Authors: Ballester Ripoll, Rafael, Leonelli, Manuele
Format: article
Publication Date:2025
Country:España
Institution:IE
Repository:Repositorio IE
OAI Identifier:oai:repositorio.ie.edu:20.500.14417/3904
Online Access:https://doi.org/10.1016/j.ijar.2025.109368
https://hdl.handle.net/20.500.14417/3904
https://www.sciencedirect.com/science/article/abs/pii/S0888613X2500009X
Access Level:Embargoed access
Keyword:33 Ciencias Tecnológicas
ODS 9 - Industria, innovación e infraestructura
Description
Summary:Traditionally, the sensitivity analysis of a Bayesian network studies the impact of individually modifying the entries of its conditional probability tables in a one-at-a-time (OAT) fashion. However, this approach fails to give a comprehensive account of each inputs' relevance, since simultaneous perturbations in two or more parameters often entail higher-order effects that cannot be captured by an OAT analysis. We propose to conduct global variance-based sensitivity analysis instead, whereby n parameters are viewed as uncertain at once and their importance is assessed jointly. Our method works by encoding the uncertainties as n additional variables of the network. To prevent the curse of dimensionality while adding these dimensions, we use low-rank tensor decomposition to break down the new potentials into smaller factors. Last, we apply the method of Sobol to the resulting network to obtain n global sensitivity indices, one for each parameter of interest. Using a benchmark array of both expert-elicited and learned Bayesian networks, we demonstrate that the Sobol indices can significantly differ from the OAT indices, thus revealing the true influence of uncertain parameters and their interactions.