| Summary: | 博士 === 國立成功大學 === 工業設計學系碩博士班 === 101 === Within the industrial design field, prediction models, which can analyze the relationship between consumers’ affective responses (CARs) and product form features (PFFs), have been studied extensively since CARs toward the product represents a mode of human-product interaction. However, despite a vast amount of literatures available on the subject, the heterogeneous nature of consumer preference patterns is often neglected so that the resulting prediction model is of less value for the real-word applications. This paper proposes a Kansei engineering framework for constructing the unified consensus affective response prediction model (CARPM) for CARs based on the concepts of consumer segmentation and information fusion. First, a fuzzy c-means (FCM) clustering is applied to separate the consumers with heterogeneous preference patterns into homogenous groups. In the target group, the relative importance of each consumer and the interaction between pairs of consumers can be determined according to the results of FCM clustering. Then, a state-of-the-art machine learning approach known as “support vector regression (SVR)” is used to construct the individual affective response prediction model (IARPM) for each consumer. These IARPMs have outperforming predictive ability of the affective responses due to the good generalization performance of the SVR algorithm. Finally, a fuzzy integral aggregation operator, namely the 2-additive Choquet integral, is employed to conjoint the IARPMs in each group by considering the relative importance and interactions of consumers in the target group to build the CARPM. According to the proposed framework, CARs toward PFFs can be predicted precisely by IAPRMs, and consumer groups can be clustered meaningfully by visualized way. Furthermore, CARs of the target groups could be gathered by considering the inherent interaction among the consumers. A case study of mobile phone is used to demonstrate the proposed approach. The results show that the proposed methodology is, an information fusion concept for handling the consumers’ evaluations of the target group, providing that mixture of experts is an alternative for dealing with CARs for product differentiation.
|
|---|
