Chaos-Based Fuzzy Regression Approach to Modeling Customer Satisfaction for Product Design

The success of a new product is very much related to the customer satisfaction level of the product. Therefore, it is important to estimate the customer satisfaction level of a new product in its design stage. Quality function deployment is commonly used to develop customer satisfaction models for product design. Relationships between customer satisfaction and design attributes are highly fuzzy and nonlinear, but these relationship characteristics cannot be captured by existing customer satisfaction models. In this paper, we propose a novel chaos-based fuzzy regression (FR) approach with which fuzzy customer satisfaction models with second- and/or higher order terms, and interaction terms can be developed. The proposed approach uses a chaos optimization algorithm to generate the polynomial structures of customer satisfaction models. Thereafter, it employs an FR method to determine the fuzzy coefficients of the individual terms of models. To illustrate and validate the proposed approach, it is applied in the development of a customer satisfaction model for a mobile phone design. Five validation tests are conducted to compare modeling results from the chaos-based FR with those from statistical regression, FR, and fuzzy least-squares regression. Results of the validation tests show that the proposed approach outperforms the other three approaches in terms of mean relative errors and variance of errors and customer satisfaction models with second- and/or higher order terms, and interaction terms can be developed effectively using the proposed chaos-based FR approach.