RECORD-TO-RECORD TRAVEL ALGORITHM FOR ATTRIBUTE REDUCTION IN ROUGH SET THEORY (original) (raw)

Attribute reduction is the process of selecting a minimal attribute subset from a problem domain while retaining a suitably high accuracy in representing the original attributes. In this work, we propose a new attribute reduction algorithm called record-to-record travel (RRT) algorithm and employ a rough set theory as a mathematical tool to evaluate the quality of the obtained solutions. RRT is an optimization algorithm that is inspired from simulated annealing, which depends on a single parameter called DEVIATION. Experimental results on 13 well known UCI datasets show that the proposed method, coded as RRTAR, is comparable with other rough set-based attribute reduction methods available in the literature.