A hybrid ANFIS-NSGAⅡ approach to modelling customer satisfaction for affective design
Author: Jiang, Huimin; Kwong, C.K.
Institution: Department of Industrial and Systems EngineeringThe Hong Kong Polytechnic UniversityHong Kong, China
Section: Interaction Design
One of the key issues in affective design is how design attribute settings of new products can be determined such that a high degree or even maximum customer satisfaction can be obtained. In this regard, modelling of the relationships between customer satisfaction and design attributes in affective design need to be performed first. Adaptive neuro-fuzzy inference systems (ANFIS) has been shown in previous research to be an effective approach to modelling affective relationships, which can deal with the fuzziness of survey data and model the nonlinear behaviour of the affective relationships. However, ANFIS is incapable of modelling the relationships that involve a number of inputs. In this paper, a hybrid ANFIS-non-dominated sorting genetic algorithmⅡ (ANFIS-NSGAⅡ) approach is proposed to overcome the limitation of ANFIS in modelling customer satisfaction for affective design. In the proposed approach, NSGAⅡ is introduced to determine the input attributes by solving a bi-objective optimization problem, in which the objectives include minimizing complexity of ANFIS and modelling errors, and then the corresponding customer satisfaction model is developed based on ANFIS. To illustrate and validate the proposed approach, a case study of affective design is conducted based on the proposed approach. The modelling results based on the proposed approach are compared with those based on ANFIS, fuzzy least-squares regression, fuzzy regression, and genetic programming-based fuzzy regression. The training and validation results show that the proposed approach outperforms the others in terms of mean absolute percentage error and variance of errors.