Rule discovery process based on rough sets under the belief function framework

Abstract

In this paper, we deal with the problem of rule discovery process based on rough sets from partially uncertain data. The uncertainty exists only in decision attribute values and is handled by the Transferable Belief Model (TBM), one interpretation of the belief function theory. To solve this problem, we propose in this uncertain environment, a new method based on a soft hybrid induction system for discovering classification rules called GDT-RS which is a hybridization of the Generalization Distribution Table and the Rough Set methodology. © 2010 Springer-Verlag Berlin Heidelberg.