Resumo

SOTO, CARLOS  e  JIMENEZ, CLAUDIA. SUPERVISED LEARNING FOR FUZZY DISCRIMINATION AND CLASSIFICATION. Dyna rev.fac.nac.minas [online]. 2011, vol.78, n.169, pp.26-33. ISSN 0012-7353.

The objective of this paper is to propose a new method of fuzzy discriminant analysis, which makes use of supervised learning strategy and uses the Euclidean distance as dissimilarity measure. The proposed formulas can solve problems of discrimination and classification of objects in categories which cannot be defined precisely because they have some overlapping degree. In order to illustrate the proposed method, we used a well-known reference database in pattern recognition. We presented the surfaces of membership functions for some examples. Additionally, the method proposed was compared with the Fisher's proposal for showing that the contour plot of 0.5 membership degree coincide with the discriminant linear model when two categories are considered. Finally, we concluded that the technique proposed is easy to implement and computationally efficient.

Palavras-chave : Fuzzy Discrimination and Classification; Fuzzy Discriminant Analysis; Supervised Learning, Fuzzy Logic; Data Mining.

        · resumo em Espanhol     · texto em Espanhol     · Espanhol ( pdf )