Comparison of Two Algorithms of Attribute Reduction Based on Fuzzy Rough Set

Abstract

Currently, with the large number of data and the increasing importance of it, how to find useful pattern in the large data, has become an important application of data mining. The rough set attribute reduction algorithm, used to study how to contain the same information when we use fewer properties to describe the objects, has been more widely used, so that the concept of soft computing is becoming increasingly popular. Rough set attribute reduction algorithm can only be applied to discrete data sets, and how to apply it to the continuous collections of the real data is a hot issue in the fuzzy mathematics. By applying the concept of fuzzy set in this issue, we can reduce the loss of information in discretization of continuous attributes. Thus the reduction results have less properties for description and contain the same information at the same time. Because of the difference between the directions of fuzzy set theory applications, that is, the reduction is based on the degree of dependence or the discernibility matrices. It can produce different fuzzy rough set attribute reductions. CCD-FRSAR(attribute reduction based on the compact computational domain of fuzzy-rough set) and FRSAR-SAT (fuzzy-rough set attribute reduction of satisfiability problem)are new and have practical values in these algorithms. Two algorithms have different ways to apply fuzzy sets theory, so the effects of them are different, too. This article describes the related ideas of fuzzy mathematics, describes the two algorithms and compares them.