Support vector machine-based optimized decision threshold adjustment strategy for classifying imbalanced data

作者：

Highlights：

• We analyze the reason why SVM can be damaged by class imbalance in theory.

• We propose SVM-OTHR algorithm to find the optimal moving distance of hyperplane.

• We integrate SVM-OTHR into Bagging ensemble framework to promote its robustness.

• The time complexity of SVM-OTHR is merely a little higher than standard SVM.

• Two proposed algorithms often outperform some other bias correction algorithms.

摘要

•We analyze the reason why SVM can be damaged by class imbalance in theory.•We propose SVM-OTHR algorithm to find the optimal moving distance of hyperplane.•We integrate SVM-OTHR into Bagging ensemble framework to promote its robustness.•The time complexity of SVM-OTHR is merely a little higher than standard SVM.•Two proposed algorithms often outperform some other bias correction algorithms.

论文关键词：Class imbalance,Support vector machine,Decision threshold adjustment,Optimization search,Ensemble learning

论文评审过程：Received 3 May 2014, Revised 4 December 2014, Accepted 5 December 2014, Available online 13 December 2014.

论文官网地址：https://doi.org/10.1016/j.knosys.2014.12.007