Generate (F, ε)-dynamic reduct using cascading hashes

Abstract

Dynamic reducts with large stability coefficients are good candidates for decision rules generation but it is time consuming to generate them. This paper presents an algorithm dReducts using a cascading hash function to generate (F, ε)-dynamic reducts. With the cascading hash function, an F-dynamic reduct can be generated in O(m 2 n) time with O(mn) space where m and n are total number of attributes and total number of instances of the table. Empirical results of generating (F, ε)-dynamic reducts using five of ten most popular UCI datasets are presented and they are compared to the Rough Set Exploration System (RSES). © 2010 Springer-Verlag Berlin Heidelberg.