Improving indiscernibility matrix based approach for attribute reduction

Abstract

The problem of finding all reducts of the attribute set of a data table has often been studied in rough set theory by using the notion of discernibility matrix. An alternative version based on the indiscernibility matrix has been used for this problem to a lesser extent due to its space and time complexity. This paper improves the indiscernibility matrix based approach for computing all reducts. Only indiscernibility matrix cells as well as subsets of the attribute set necessary for computing reducts are processed. The experiments reported in this paper show that the improved version uses less memory to store partial results and can find reducts in a shorter time.