Discriminant analysis of fuzzy data and its application to gas sensor system

In the conventional discriminant analysis, attribute values of each sample are represented by a point in a multi-dimensional feature space. In this paper, the authors propose a discriminant method for fuzzy data where attribute values of each sample are represented by a fuzzy number vector in the feature space. The proposed method is based on a fuzzy linear discriminant function which maps the fuzzy number vector into a fuzzy number in a one-dimensional decision space. A mathematical programming program, which can be solved using the linear programming technique, is formulated to determine the coefficients of the discriminant function. For the application of the proposed method, the authors discriminate the fuzzy response patterns of nine different gases measured by nine different gas sensors. [In Japanese.]