Symbolic Representation for Rough Set Attribute Reduction Using Ordered Binary Decision Diagrams

Abstract

The theory of rough set is the current research focus for knowledge discovery, attribute reduction is one of crucial problem in rough set theory. Most existing attribute reduction algorithms are based on algebra and information representations, discernibility matrix is a common knowledge representation for attribute reduction. As problem solving under different knowledge representations corresponding to different difficulties, by changing the method of knowledge representation, a novel knowledge representation to represent the discernibility matrix using ordered binary decision diagrams (OBDD) is proposed in this paper, the procedures to translate the discernibility matrix model to the conversion OBDD model is presented, experiment is carried to compare the storage space of discernibility matrix with that of OBDD, results show that OBDD model has better storage performance and improve the attribute reduction for those information systems with more objects and attributes, it provide the foundation for seeking new efficient algorithm of attribute reduction. Index Terms—rough set, attribute reduction, discernibility matrix, ordered binary decision diagrams