A fuzzy classifier that uses both crisp samples and linguistic knowledge

We propose a general structure for a classifier that uses fuzzy inference techniques. This structure leads to a variety of fuzzy logic classifiers that are capable of combining numerical data and linguistic knowledge in a unified framework. Under certain conditions the fuzzy logic classifier reduces to the Bayes minimum error classifier. Our examples show that, when linguistic information is available, the fuzzy classifiers can perform better than probabilistic classifiers that do not use the linguistic information