Apply Multi-class Fuzzy Support Vector Machines to Product-Form-Image Prediction

摘要

Extracting critical form features from a product relative to some specific Kansei adjectives through Kansei image questionnaire system is an important method. Due to the fuzziness of Kansei image, a fuzzy classifier needs to be constructed. In this study, a kansei evaluation on product form employing multi-class fuzzy support vector machines (MF-SVMs) was proposed to extract implicit information of products. Critical form features were mapped into an n-dimensional vector; for multi-dimensional Kansei images, "One-Versus-Rest" (OVR) method for multi-class SVMs was addressed to deal with this problem. For new products, system will specify it by using MF-SVMs based classification model. A case study of mobile phone design is given to demonstrate the effectiveness of the proposed methodology.