Model-based decision support systems are preferred due to consistency in decision-making and due to time-efficiency in model evaluation and modification. Constructing a model may take time if a number of random variables are involved in the model and the model structure is not simple. Classification is a form of decision making under a certain loss structure.
Consider a decision support system based on a Bayesian network(BN) where all the variables involved are binary, each taking on 0 or 1. The system categorizes the probability that a certain variable is equal to 1 conditional on a set of variables in an ascending order of the probability values and predicts for the variable in terms of category levels.
We introduce the notion of similarity between BN models and propose a method of constructing a BN whose prediction is robust when the prediction is made for a variable in terms of category levels of probability. We have considered a uniform distribution, a beta distribution, and a variation of the beta distribution for the distributions of the conditional probabilities. We investigate how the agreement levels are affected in a BN model by simulation. We also explored the distribution of the total score of a test, and we showed that the total score has a normal distribution under some conditions.