An Over-sampling Method Based on Margin Theory

Abstract

Imbalanced data widely exists in real life, while the traditional classification method usually takes accuracy as the classification criterion, which is not suitable for the classification of imbalanced data. Resampling is an important method to deal with imbalanced data classification. In this paper, a margin based random over-sampling (MRO) method is proposed, and then MROBoost algorithm is proposed by combining the AdaBoost algorithm. Experimental results on the UCI dataset show that the MROBoost algorithm is superior to AdaBoost for imbalanced data classification problem.