Knowledge granularity reduction for decision tables

Abstract

Attribute reduction is a difficult topic in rough set theory and knowledge granularity reduction is one of the important types of reduction. However, up to now, its reduction algorithm based on a discernibility matrix has not been given. In this paper, we show that knowledge granularity reduction is equivalent to both positive region reduction and X-absolute reduction, and derive its corresponding algorithm based on a discernibility matrix to fill the gap. Particularly, knowledge granularity reduction is the usual positive region reduction for consistent decision tables. Finally, we provide a simple knowledge granularity reduction algorithm for finding a reduct with the help of binary integer programming, and consider six UCI datasets to illustrate our algorithms.