ACQUISITION OF KANSEI DECISION RULES OF COFFEE FLAVOR USING ROUGH SET METHOD
Tatsuo Nishino, Mitsuo Nagamachi, Masatoshi Sakawa · KANSEI Engineering International · 2006
The acquisition of effective decision rules between design elements of products and human evaluations is significant, but difficult especially in the cases where we have to handle linearly inseparable and much ambiguous data. We have proposed the rough set method to handle these cases. In this paper, we also propose the utilization of gain chart that can visualize a relation between the effects of product attributes on decision and the number of evaluation events. Using the gain chart, we can set up a suitable parameter corresponding with gain property of each evaluation word. We show an application of our proposed rough set method to the extraction of decision rules between coffee flavor evaluations, i.e., taste and aroma, and the pattern of coffee manufacturing conditions. The results showed that our rough set method enabled more effectively to extract the combination rules of manufacturing conditions from much ambiguous data like sense and feeling. We also found out that it is easy to obtain more general rules by β-upper approximation as well as β-lower approximation.