On the Comparative Study of Prediction Accuracy for Credit Card Fraud Detection wWith Imbalanced Classifications

Abstract

Credit card fraud is one of the critical issues due to its significant losses to both financial institutions and individuals in the society. The accurate detection and prevention of fraudulent activities are necessary to protect financial institutions and individuals. This paper performs a comparative experimental study to detect credit card frauds, as well as to tackle the imbalance classification problem by applying different machine learning algorithms for handling imbalanced datasets. Our study shows that there is no need to process imbalance dataset by applying resampling techniques to measure the performance of our classifiers and it is sufficient to measure the performance through the three-performance measurements (Accuracy, Sensitivity, and Area Under Precision/Recall Curve (PRC)) to prove the accuracy of the prediction of classification. Finally, by handling imbalanced classifications with imbalance datasets, one can minimize the number of false alarms, reduce damages to financial institutions and individuals, increase accurate for fraud detection, and decrease the occurrence of fraud cases using machine learning techniques.