Class Decomposition via Clustering: A New Framework for Low-Variance Classifiers

Abstract

We propose a pre-processing step to classification that applies a clustering algorithm to the training set to discover local patterns in the attribute or input space. We demonstrate how this knowledge can be exploited to enhance the predictive accuracy of simple classifiers. Our focus is mainly on classifiers characterized by high bias but low variance (e.g… (More)
DOI: 10.1109/ICDM.2003.1251005

Topics

3 Figures and Tables

Cite this paper

@inproceedings{Vilalta2003ClassDV, title={Class Decomposition via Clustering: A New Framework for Low-Variance Classifiers}, author={Ricardo Vilalta and Murali-Krishna Achari and Christoph F. Eick}, booktitle={ICDM}, year={2003} }