A Review of Machine Learning Applications for Credit Card Fraud Detection with A Case study

Purpose - This paper aims to highlight the widely used supervised techniques applied for fraud detection. In addition, this paper aims to apply some techniques to evaluate their performance on real-world data and develop an ensemble model as a potential solution for this problem. Design/Methodology- Different techniques applied in this study for fraud detection purposes are logistic regression, decision tree, random forest, KNN, and XGBoost. The confusion matrix gives information about the assignment of inputs to the different classes. This study uses precision and recall to evaluate the performance, calculated based on the confusion matrix. Findings- XGBoost is the fastest and is expected to have the best performance; however, it is only outperforming the random forest in terms of accuracy, precision, recall, and f1-score. In general, the KNN and logistic regression have better performance, which means they better detect fraudulent transactions. Practical Implications- The new model can be applied to new data instead of the previous techniques.