A comparative study of fuzzy and neural network approaches to discriminant analysis with linguistic variables

Abstract

This paper proposes a fuzzy discriminant analysis to solve the two-group classification problem where the measured variables are linguistic in nature. Especially under imprecise framework, the linguistic variables capture more information although vagueness is inherent. In analogy to classical statistics, a fuzzy linear discriminant function is introduced here, which directly deals with continuous fuzzy numbers as the representative of linguistic values to obtain fuzzy scores for classification. To make a comparative study, the backpropagation neural network approach has also been studied in this paper. Finally admission to management programme is considered as an example of the application on two-level classification problem of the proposed method.