Credit card fraud detection: an evaluation of SMOTE resampling and machine learning model performance

Credit card fraud has been a noted security issue that requires financial organisations to continuously improve their fraud detection system. In most cases, a credit transaction dataset is expected to have a significantly larger number of normal transactions than fraud transactions. Therefore, the accuracy of a fraud detection system depends on building a model that can adequately handle such an imbalanced dataset. The purpose of this paper is to explore one of the techniques of dataset rebalancing, the synthetic minority oversampling technique (SMOTE). To evaluate the effects of this technique on model training, we selected four basic classification algorithms, complement naïve Bayes (CNB), K-nearest neighbour (KNN), random forest and support vector machine (SVM). We then compared the performances of the four models trained on the rebalanced and original dataset using the area under precision-recall curve (AUPRC) plots.