Bayesian networks describe the nature of the probabilistic inter-dependence of factors (or ‘nodes’) in an influence diagram. Conditional probabilities for complex belief networks can be calculated by computer algorithms using Bayes Theorem. The model evolves as new information is collected, so that the model constantly reflects the current state of knowledge about the system. Bayesian networks have been used extensively to model real world problems. A number of commercially available software packages exist, which enable networks to be constructed using standard desktop computers. (click to enlarge)

Key Ideas

• Bayesian networks are a way to model aspects of a system when certain conditional probabilities are known or can be estimated
• Bayesian networks may help develop values for consequence tables once objectives and alternatives have been defined