Fuzzy classifier based on fuzzy support vector machine

Abstract

Support vector machines SVMs have been very successful in pattern recognition and function estimation problems. When SVMs are used for classification, the inputs of the training example are real-valued and the outputs are class label y = ±1. However, in practice, the training examples usually belong to a class with certain fuzzy membership, therefore it is important to consider uncertain class label for classification problems. For this purpose, this paper introduces the new concept of fuzzy hyperplane, and constructs the fuzzy classifiers based on fuzzy support vector machines. At the end of the paper, we apply our new methods to medical diagnosis problems.