MASI: Moving to adaptive samples in imbalanced credit card dataset for classification

Abstract

Fraud in financial areas is broadly going to cause significant consequences, in recently. As a result, financial fraud detection is interested in many researchers. The imbalanced dataset in classification might influence to the prediction results as its bias. In this paper, an improvement algorithm, so-called as MASI, is proposed for financial fraud detection in imbalanced data classification. The experiment is performed on UCI machine learning repository data domain. Our results show the better in sensitivity, specificity, and G-mean values compared to other control methods such as Random Over-sampling, Random Under-sampling, SMOTE and Borderline SMOTE in using classification algorithms (SVM, C50 and RF).