The Bayesian Network quality is measured depending on your problem.
All variables are equally important (unsupervised learning
and clustering): we score how much the Bayesian Network fits the
data.
Knowledge focuses on one particular variable as in supervised learning,
then quality is evaluated by a confusion matrices (occurences, precision and
reliability matrices) and lift and ROC curves.
BayesiaLab
Evaluation
All variables are equally important (unsupervised learning
and clustering): we score how much the Bayesian Network fits the
data.