Affiliations: [a] Department of Electrical and Computer Engineering, Curtin University, Australia
| [b] Faculty of Engineering and Information Technology, University of Technology, Sydney, Australia
Correspondence:
[*]
Corresponding author. Kit Yan Chan, Department of Electrical and Computer Engineering, Curtin University, Australia. Tel.: +61 8 92662945; E-mail: [email protected].
Abstract: Fuzzy regression models have commonly been used to correlate engineering characteristics with consumer preferences regarding a new product. Based on the models, product developers can determine optimal engineering characteristics of the new product in order to satisfy consumer preferences. However, they have a common limitation in that they cannot guarantee to include significant regressors with significant engineering characteristics or significant nonlinear terms. The generalization capability of the model can be reduced, when too few significant regressors are included and too many insignificant regressors are included. In this paper, a forward selection based fuzzy regression (FS-FR) is proposed based on the statistical forward selection to determine significant regressors. After the significant regressors are determined, the fuzzy regression is used to generate the fuzzy coefficients which address the uncertainties due to fuzziness and randomness caused by consumer preference evaluations. The developed model includes only significant regressors which attempt to improve the generalization capability. A case study of a tea maker design demonstrated that the FS-FR was able to generate consumer preference models with better generalization capabilities than the other tested fuzzy regressions. Also simpler consumer preference models can be provided for the new product development.
