We propose expected attainable discrimination (EAD) as a measure to se
lect discrete valued features for reliable discrimination between two
classes of data. EAD is an average of the area under the ROC curves ob
tained when a simple histogram probability density model is trained an
d tested on many random partitions of a data set, EAD can be incorpora
ted into various stepwise search methods to determine promising subset
s of features, particularly when misclassification costs are difficult
or impossible to specify. Experimental application to the problem of
risk prediction in pregnancy is described. (C) 1998 Elsevier Science B
,V, All rights reserved.