Abstract
Expert opinion is increasingly being used to inform Bayesian Belief Networks, in particular to define the conditional dependencies modelled by the graphical structure. The elicitation of such expert opinion remains a major challenge due to both the quantity of information required and the ability of experts to quantify subjective beliefs effectively. In this work, we introduce a method designed to initialise conditional probability tables based on a small number of simple questions that capture the overall shape of a conditional probability distribution before enabling the expert to refine their results in an efficient way. These methods have been incorporated into a software Application for Conditional probability Elicitation (ACE), freely available at https://github.com/KirstyLHassall/ACE (Hassall, 2019).
Highlights
Bayesian Belief Nets, referred to as Bayes Nets, Belief Networks or often BBNs, have, in recent years, seen a dramatic increase in their use for describing and modelling natural systems
We focus on the key issues surrounding the use of expert opinion in the characterisation of conditional probability tables
Group A were given 25 minutes to complete as many conditional probability table (CPT) of the road safety network as they could using the scoring system described above. They were given 25 minutes to fill in as many CPTs without using the scoring system, i.e. to fill in the tables manually, a graphical aid remained accessible in the software
Summary
Bayesian Belief Nets, referred to as Bayes Nets, Belief Networks or often BBNs, have, in recent years, seen a dramatic increase in their use for describing and modelling natural systems. Examples include quantifying the risk of erosion in peat bogs (Aalders et al, 2011), modelling ecosystem services (Haines-Young, 2011), applications in natural resource management (see Henriksen et al, 2012, and references therein), mapping risks of soil threats such as soil compaction (Troldborg et al, 2013), predicting soil bulk density at landscape scales (Taalab et al, 2015) and assessing the impact of buffer zones on water protection and biodiversity (Tattari et al, 2003) This explosion in practical BBN modelling may in part be due to the relative simplicity of the intuitive graphical representation of multiple interrelated variables captured through conditional probabilities and more practically, the increasing accessibility to specialist BBN software. We focus on the key issues surrounding the use of expert opinion in the characterisation of conditional probability tables.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
