Rough sets is a mathematically well-grounded approach to discovering p
atterns in data. Examples related to materials science illustrate the
application of rough sets methodology to identifying related objects,
determining the degree of dependency of subsets of attributes on other
attributes, and testing subsets of objects for patterns. Classificati
on of objects into classes that exhibit specific behavior is important
in understanding material behavior and in predicting material propert
ies. The lower approximation identifies those objects that belong with
certainty to a pattern; the upper approximation identifies those obje
cts that have some probability of belonging to some pattern. Results o
f the application of the rough sets method to the chalcopyrite family
of crystal structures are presented. Copyright (C) 1996 Acta Metallurg
ica Inc.