Abstract
We discuss value reduction in rough set theory based on comparison between partitions on the universe of discourse. In rough set theory, value reduction is as important for generating concise decision rules from decision tables as attribute reduction. In this paper, we point out that comparison between partitions generated by values of condition attributes and decision attributes in the given decision table and gathering partitions generated by the values of condition attributes as small number of values as possible to describe the positive region of decision classes corresponds to value reduction of condition attributes. We also propose a heuristic value reduction algorithm based on the above idea and evaluate the usefulness of the proposed algorithm
by experiments.
