Title: KNN-based ensemble selection for imbalance learning

Authors: Guirong Zheng; Chang-an Wu; Huaping Guo

Addresses: School of Computer and Information Technology, Xinyang Normal University, Xinyang, 464000, China ' School of Computer and Information Technology, Xinyang Normal University, Xinyang, 464000, China ' School of Computer and Information Technology, Xinyang Normal University, Xinyang, 464000, China

Abstract: Classification of imbalance datasets is one of the crucial issues in the field of machine learning. Since the distribution of imbalance dataset is extremely skew, the traditional classifications often come up with a disappointed performance. Different with the traditional methods, this paper reconsiders class imbalance problem from the viewpoint of ensemble learning. However, many ensembles tend to build base classifiers not perfectly; some bad performed classifiers which are helpless to improve the generalised ability of ensemble would be produced. To solve this problem, a novel method named ensemble algorithm named NNES (k-nearest neighbour based ensemble section) was proposed in this paper. To evaluate the local properties of an unlabeled instance, NNES tends to pay more attention to minority and improve its performance on imbalance datasets. Experimental results show that NNES can improve the classification performance of the imbalance datasets effectively. Moreover, this improvement would be strengthened when the sampling techniques were introduced in.

Keywords: imbalance datasets; ensemble; k-nearest neighbour; classification; sampling techniques.

DOI: 10.1504/IJCSYSE.2019.100025

International Journal of Computational Systems Engineering, 2019 Vol.5 No.2, pp.82 - 96

Received: 08 Mar 2017
Accepted: 04 Oct 2017
Published online: 04 Jun 2019 *

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