Benefits of Bayesian Network Models

2Bayesian Network: Modeling Formalism of the Structure Function of Boolean Systems

2.1. Introduction

One of the principle characteristics of modeling Boolean stucture function using BN is the ability to construct models from knowledge without technical expertise regarding computing algorithms. Nevertheless, this advantage can be a source of doubt about the computing results obtained from BN models. Formally, the numerical results are exact and the question of validity should concern only the quality of the model built by the analyst and/or the representativeness of data used to learn the parameters. Therefore, it is very important to use a structured modeling approach to obtain a model that better corresponds to reality.

From a practical point of view, there is often a lack of data to inform models in reliability estimation, risk analysis and maintenance optimization. It is often impossible to fully define the joint distribution defining all situations and their associated probabilities. As a result, modeling tools require the use of expert judgments to build structured models [CEL 06]. The BN modeling practiced in this book is presented in this spirit.

BN is a powerful modeling tool as it can combine knowledge of different kinds. This combination is allowed by the probabilistic representation and the combination of state of affairs. The model structure as well as the estimation of the model parameters can be built either automatically or manually from: data from feedback experiences; ...