A comparison of re-sampling techniques for pattern classification in imbalanced data-sets

Authors: Saul, M.A. and Rostami, S.

http://eprints.bournemouth.ac.uk/31059/

Start date: 5 September 2018

This data was imported from Scopus:

Authors: Saul, M.A. and Rostami, S.

http://eprints.bournemouth.ac.uk/31059/

Journal: Advances in Intelligent Systems and Computing

Volume: 840

Pages: 240-251

ISBN: 9783319979816

ISSN: 2194-5357

DOI: 10.1007/978-3-319-97982-3_20

© Springer Nature Switzerland AG 2019. Class imbalance is a common challenge when dealing with pattern classification of real-world medical data-sets. An effective counter-measure typically used is a method known as re-sampling. In this paper we implement an ANN with different re-sampling techniques to subsequently compare and evaluate the performances. Re-sampling strategies included a control, under-sampling, over-sampling, and a combination of the two. We found that over-sampling and the combination of under- and over-sampling both led to a significantly superior classifier performance compared to under-sampling only in correctly predicting labelled classes.

The data on this page was last updated at 04:51 on December 9, 2018.