SCHOOL OF COMPUTER SCIENCE AND SOFTWARE ENGINEERING
MONASH UNIVERSITY


TECHNICAL REPORT 2005/170


Knowledge Engineering Cardiovascular Bayesian Networks from the Literature

C R Twardy, A E Nicholson and K B Korb

ABSTRACT

Bayesian networks are rapidly becoming a tool of choice for applied Artificial Intelligence. There have been many medical applications of BNs however few applying data mining methods to epidemiology. In a previous study we looked at such an application to epidemiological data, specifically assessment of risk for coronary heart disease. . In that previous study, we featured two Bayesian networks "knowledge-engineered" from the epidemilogy literature, but postponed a detailed discussion of their construction. This report provides the full details of our engineering choices, and the reasons for them. It will interest anyone wishing to replicate our results, or check our assumptions or methods. It may also be of interest to others wishing to make a similar Bayesian network from the epidemiological literature for risk prediction of other medical conditions, as it provides a case study in the steps that need to be undertaken. We used the Bayesian network software package Netica to implement the BNs which generated particular changes. This report notes specific Netica traps and tricks, which may help others avoid some of the difficulties we encountered. However, the approach described here should extend to any system allowing equations to specify the probability distributions for each node.