Hi everybody, I want to know if there exist some kind of consistency measurement for Bayesian Networks if the probability tables were made up from expert knowledge, that is, there is now databases involved. Maybe there is some sensitivity analysis that I can use? I have also think about the gain theory that is used in Decision Trees, so that I can examine the gain that is obtained from each parent node to a child node and also measure the gain if the probabilities of the parent nodes are changed. This can then also be used as a measure of consistency to the whole Bayesian Network. I don't know Decision Trees that well, though and I am afraid of making a fundamental error. I am reading the following article: A Bayesian Approach to Learning Bayesian Networks with Local Structure, DM Chickering, D Heckerman, C Meek.; Technical Report MSR-TR-97-07, (1997) The article discuss the use of Decision tree representation for the probability distributions. Maybe someone with knowledge of both fields (Bayesian Networks and Desicions Trees) can give me some advice on how to tackle this problem. Thank you very much for reading this e-mail Greetings Alta de Waal Modelling & Simulation Defence Electronics CSIR tel: +27 (12) 841 3792 www.csir.co.za
