Abstract
In this paper, we discuss a fuzzy classifier with polyhedral regions. First, for each class we generate a hyperbox calculating the minimum and maximum values of the data belonging to the class. Next, we cut the hyperbox using training data belonging to the other classes so that class separability is maximized. Finally, for each convex polyhedron we define a membership function using the minimum operator. We demonstrate the superiority of our method over our previously developed classifier with polyhedral regions using thyroid, numeral, hiragana, and blood cell data sets.
