A New Fuzzy Discriminant Analysis Method

摘要

A new more informative and effective fuzzy discriminant analysis method based on fuzzy regression with point prototypes has been developed and applied on two relevant data sets (the classical Fisher's Iris data set and a clinical data set concerning different diseases). The proposed fuzzy method is consistent with the supervised character of the original discriminant analysis method. The classification and patterns obtained by membership degrees plot are in a very good agreement with the structure of data and the initial assignment of samples, which indicate that the new approach may be successfully employed in different fields. In addition, the graphical representation of fuzzy membership degrees to different classes provides a relevant way to visualize the relationships between the data items of the fuzzy classes.