Existing data sets of cases can significantly reduce the knowledge engineer
ing effort required to parameterize Bayesian networks. Unfortunately, when
a data set is small, many conditioning cases are represented by too few or
no data records and they do not offer sufficient basis for learning conditi
onal probability distributions, We propose a method that uses Noisy-OR gate
s to reduce the data requirements in learning conditional probabilities. We
test our method on HEPAR II, a model for diagnosis of liver disorders, who
se parameters are extracted from a real, small set of patient records. Diag
nostic accuracy of the multiple-disorder model enhanced with the Noisy-OR p
arameters was 6.7% better than the accuracy of the plain multiple-disorder
model and 14.3% better than a single-disorder diagnosis model. (C) 2001 Els
evier Science Inc. All rights reserved.