A NEURO-FUZZY APPROACH TO DATA-ANALYSIS OF PAIRWISE COMPARISONS

Artificial neural networks provide iterative on-line learning schemes for modeling non-linear systems. An iterative learning algorithm in fuzzy models, which is called neuro-fuzzy, has been recently developed within the framework of fuzzy modeling in the sense of M. Sugeno. In this paper, using neuro-fuzzy approach, two quantification methods of pairwise comparisons are presented in order to derive the associated weights of different objects. The proposed methods can be applied even in the case of incomplete pairwise comparisons. A simplified fuzzy reasoning model is obtained in the form of Gaussian radial basis functions. The psychological sensation responses of human beings to minute vibrations are analyzed by the newly proposed neuro-fuzzy approach. The proposed approach is compared with Guttman's method and Saaty's analytic hierarchy process (AHP). In our two neuro-fuzzy approaches, psychological values are obtained with the interval and the ratio scale properties. They are represented by smooth functions of class C(infinity).