Improving k-NN by using fuzzy similarity functions

Abstract

The k-Nearest Neighbor (k-NN) is the basis for many lazy learning algorithms. It uses a similarity function to generate predictions from stored instances. Many authors have shown that the performance of k-NN is highly sensitive to the definition of its metric similarity. In this paper we propose the use of the fuzzy set theory in the definition of similarity functions. We present experiments with a real application to demonstrate the usefulness of this approach. © Springer-Verlag Berlin Heidelberg 2004.