An application of fuzzy support vectors

Abstract

Support Vector Machines (SVMs) are a recently introduced Machine Learning technique. SVMs approach binary classification by attempting to find a hyperplane that separates the two categories of training vectors. This hyperplane is expressed as a function of a subset of the training vectors. These vectors are called support vectors. In this paper, we present a method of fuzzifying support vectors based off of the results of an SVM induction. We then propose a method of enhancing SVM induction using these fuzzy support vectors. We finish by presenting a computational example using the IRIS data set.