Ranked nodes: A simple and effective way to model qualitative judgements in large-scale Bayesian Networks

Ranked nodes: A simple and effective way to model qualitative judgements in large-scale Bayesian Networks Norman Fenton and Martin Neil Risk Assessment and Decision Analysis Research Group Department of Computer Science, Queen Mary, University of London and Agena Ltd 21 Feb, 2005 Abstract Although Bayesian Nets (BNs) are increasingly being used to solve real world risk problems, their use is still constrained by the difficulty of constructing the node probability tables (NPTs) for each node. In the absence of hard data, we must rely on domain experts to provide, often subjective, judgements to inform the NPTs. A key challenge is to construct relevant NPTs using the minimal amount of expert elicitation, recognising that it is rarely cost-effective to elicit complete sets of probability values. We describe a simple approach to defining NPTs for a large class of commonly occurring nodes (called ranked nodes). The approach is based on the doubly truncated Normal distribution with a central tendency that is invariably a type of weighted function of the parent nodes. In extensive real-world case studies we have found that this approach is sufficient for generating the NPTs of a very large class of nodes. The approach has been automated and is thus accessible to all types of domain experts, including those with little statistical expertise. The result has been that such individuals have been able to build large-scale realistic BN models that solve important problems. Hence, this work represents a breakthrough in BN research and technology since it can make the difference between being able to build realistic BN models and not.