Fuzzy decision trees by fuzzy ID3 algorithm and its application todiagnosis systems

摘要

A popular and particularly efficient method for making a decisiontree for classification from symbolic data is ID3 algorithm. Revisedalgorithms for numerical data have been proposed, some of which divide anumerical range into several intervals or fuzzy intervals. Theirdecision trees, however, are not easy to understand. We propose a newversion of ID3 algorithm to generate an understandable fuzzy decisiontree using fuzzy sets defined by a user. We apply it to diagnosis forpotential transformers by analyzing gas in oil