Study on Characteristics of the Clustered Groups based on Individuality

Abstract

Recently, companies often develop marketing strategy such as the discovery of buying groups for the target products and the prediction of the market size by analyzing Kansei data such as questionnaire. In conventional analyses, subjects are classified based on the attributes such as gender, age and so on, and then the classified groups are analyzed or compared one another. On the other hand, the authors have proposed the clustering methods based on "individuality", which is the character or tendency of each subject. One of the proposed methods can extract "Kyokan group", of which the subjects favor, or moreover get impressed with the evaluation objects. It is, however, important for the development of market strategy to grasp the characteristics of the subjects in Kyokan group, i.e. the attributes of them. This paper proposes a new method to extract effective combination of attributes which shows the characteristics of "Kyokan Group" from a lot of candidates by using Genetic Algorithm. The author applies the proposed method to Web questionnaire data and studies the effectiveness of this method.