Multiclass SVM-RFE for product form feature selection

Various form features affect Consumer preference regarding product design. It is, therefore, important that designers identify these critical form features to aid them in developing appealing products. However, the problems inherent in choosing product form features have not yet been intensively investigated. In this paper, all approach based oil multiclass support vector machine recursive feature elimination (SVM-RFE) is proposed to streamline the selection of optimum product form features. First, a one-versus-one (OVO) multiclass fuzzy support vector machines (multiclass fuzzy SVM) model using a Gaussian kernel was constructed based on product samples from mobile phones. Second, all optimal training model parameter set was determined using two-step cross-validation. Finally, a multiclass SVM-RFE process was applied to select critical form features by either using overall ranking or class-specific ranking. The weight distribution of cacti iterative step call be used to analyze the relative importance of each of the form features. The results of our experiment show that the multiclass SVM-RFE process is not only very useful for identifying critical form features with minimum generalization errors but also call be used to select the smallest feature subset for building a prediction model with a given discrimination capability. (c) 2007 Elsevier Ltd. All rights reserved.