A learning strategy for highly imbalanced classification

Authors:
Tong Liu;Yongquan Liang;Weijian Ni
Affiliations:
Shandong university of Science and Technology, Tai'an Shandong, P. R. China;Shandong university of Science and Technology, Qingdao, P. R. China;Shandong university of Science and Technology, Qingdao, P. R. China
Venue:
Proceedings of the Third International Conference on Internet Multimedia Computing and Service
Year:
2011

Citing 8
Cited 1

A Tutorial on Support Vector Machines for Pattern Recognition

Data Mining and Knowledge Discovery
Editorial: special issue on learning from imbalanced data sets

ACM SIGKDD Explorations Newsletter - Special issue on learning from imbalanced datasets
Mining with rarity: a unifying framework

ACM SIGKDD Explorations Newsletter - Special issue on learning from imbalanced datasets
KBA: Kernel Boundary Alignment Considering Imbalanced Data Distribution

IEEE Transactions on Knowledge and Data Engineering
The relationship between Precision-Recall and ROC curves

ICML '06 Proceedings of the 23rd international conference on Machine learning
On multi-class cost-sensitive learning

AAAI'06 Proceedings of the 21st national conference on Artificial intelligence - Volume 1
SMOTE: synthetic minority over-sampling technique

Journal of Artificial Intelligence Research
A learning method for the class imbalance problem with medical data sets

Computers in Biology and Medicine

Editorial: Parameter-free classification in multi-class imbalanced data sets

Data & Knowledge Engineering

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper describes a new learning strategy on the problem of classification on overlapped and imbalanced training set. We devise an adaptive scheme for minority generating; with data cleaning of majority, new clusters are drawn to increasingly focus on the combination of new minority samples. Inspired by the essence of SVM, we extract the most informative SVs for training. An empirical study compares the performance of our strategy against traditional classification approaches on the benchmark data sets, and experimental results show that our learning strategy not only inherent data distribution, but also improve classification effectiveness and efficiency.