Applying Indiscernibility Attribute to Attribute Reduction Based on Discernibility Matrix

Abstract

Attribute reduction is one of the key problems in rough set theory, and many algorithms based on discernibility matrix have been proposed and studied about it. In order to reduce the computational complexity of discernibility matrix method, a fast counting sort algorithm is first introduced for dealing with redundant and inconsistent data in decision tables. Then, the improved discernibility matrix is presented for deleting a great number of empty elements in the classical algorithms. Finally, the minimal indiscerniblity attribute is applied to generate smaller discernibility matrix and a new attribute reduction algorithm is proposed. Experiments show that our algorithm outperforms other attribute reduction algorithms.