Using supervised clustering to enhance classifiers

Abstract

This paper centers on a novel data mining technique we term supervised clustering. Unlike traditional clustering, supervised clustering is applied to classified examples and has the goal of identifying class-uniform clusters that have a high probability density. This paper focuses on how data mining techniques in general, and classification techniques in particular, can benefit from knowledge obtained through supervised clustering. We discuss how better nearest neighbor classifiers can be constructed with the knowledge generated by supervised clustering, and provide experimental evidence that they are more efficient and more accurate than a traditional 1-nearest-neighbor classifier. Finally, we demonstrate how supervised clustering can be used to enhance simple classifiers. © Springer-Verlag Berlin Heidelberg 2005.