Title: Local reweight wrapper for the problem of imbalance

Authors: S.B. Kotsiantis

Addresses: Educational Software Development Laboratory, Department of Mathematics, University of Patras, Hellas

Abstract: Many real-world data sets exhibit skewed class distributions in which the classes are not approximately equally represented. Traditional machine learning algorithms can be biased towards majority class due to over-prevalence. This paper firstly provides a systematic study of the various methodologies that have tried to handle this problem. Finally, it presents an experimental study of these methodologies with a proposed local wrapper for reweighting training instances and it concludes that such a framework can be a more effective solution to the problem. Our method seems to allow improved identification of difficult small classes in predictive analysis, while keeping the classification ability of the other classes at an acceptable level.

Keywords: supervised machine learning; imbalanced data sets; local learning; skewed class distributions; local wrapper; reweighting training; predictive analysis; classification.

DOI: 10.1504/IJAISC.2008.021262

International Journal of Artificial Intelligence and Soft Computing, 2008 Vol.1 No.1, pp.25 - 38

Published online: 14 Nov 2008 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article