Feature Analysis of Roof Shapes Using Rough Sets Theory

The aim of this paper is to develop an optimum design method for roof shapes that satisfy both aesthetic and dynamic conditions by using the rough sets theory. This optimum system consists of the KANSEI evaluation system and the shape evaluation system. In the KANSEI evaluation system, the KANSEI questionnaire system is executed first, and the obtained KANSEI evaluation values are used as KANSEI learning. In the previous paper, a neural network was used for KANSEI learning. However, the relationship between the roof shapes and the KANSEI evaluation was unclear because the results of neural network had black box. To solve this problem, the rough sets theory was used instead of a neural network. When using the rough sets theory, many combination patterns of the decision and condition attributes must be considered; this is the reason why the estimation accuracy changes according to the combination patterns. Therefore, a number of combination patterns were generated and the estimation accuracy for each of these patterns was calculated using the column score, and the KANSEI features were analyzed. If the features between the roof shapes and the KANSEI evaluations are analyzed by using the rough sets theory, KANSEI evaluations become possible even if the roof shapes are not the same conditions.