Kansei Knowledge Extraction as Measure of Structural Heterogeneity

Representative measurements way of affective attributes is the Semantic Differential (SD) method in Kansei evaluation experiments. Structural heterogeneity of the subjective evaluations indicates the heterogeneous evaluation structure of each evaluator, which is presented individually by the specific selected factors. The objective of the structural heterogeneity modeling is not only to extract the general trends but also to identify the diverse individual evaluation structure. In this paper, we propose a Hierarchical Bayes Regression model with Heterogeneous Variable Selection (HBRwHVS), which simultaneously analyses the individual models and identifies the influential explanatory variables based on the framework of the Hierarchical Bayes Regression Modeling. The results offer the relational data between the evaluators and the selected items as carefully chosen explanatory variables. We apply the proposed method to the analysis of sensibility subjective evaluation data on traditional craft and show its effectiveness. After obtaining the estimated values of model parameters, cluster analysis is performed on subjects with the similar evaluation structures as another example of its applications.