A COMPARATIVE STUDY ON APPROXIMATIONS OF DECISION CLASS AND RULE ACQUISITION BY ROUGH SETS MODEL

This paper focuses on comparison of approximation methods in rough set method. It has been shown that rough set approach is more effective in extracting the design decision rules between human evaluation and product attributes. Since kansei evaluations to products include the ambiguity of decisions, we have proposed probabilistic rough set model based on upper approximation. In this paper, we propose β-method using dual β-lower and upper approximations to extract effective decision rules for kansei design of products. We compared the effectiveness of β-method with m-method as a lower approximation often used in Kansei Engineering applications and with 0.5-method as upper approximation in terms of the certainty, coverage and Michalski QM measures of decision rule through a practical application to the extraction of design decision rules for children shoes, so called toddler, Generally, the result showed that β-method using dual β-lower and upper approximations is more effective in extracting ‘interesting’ decision rules for product design as well as general ones compared with the other approximations.