Abstract
For realization of a fuzzy inference, it is necessary to give appropriate fuzzy rules. Some methods which get these fuzzy rules are proposed. A fuzzy inductive learning algorithm is one of it and gets fuzzy rules from a training example set, where a training example consists of some attributes and a discrete class. However, this algorithm can't directly deal with a continuous class. If the fuzzy rules want to deal with a continuous class, it is necessary to give a relation with a continuous class and a discrete class. In this paper, I propose the method which deals with the continuous class and a revised inductive learning algorithm composed with the method. This algorithm automatically transforms a example with a continuous class into a example with a discrete class, learns fuzzy rules which deal with the discrete class, and transforms the rules which deal with the continuous class. Also, I examine the efficiency of the algorithm by two numerical experiments.
