Credit card fraud detection and risk management strategies: A deep learning-based approach for EU banks

Abstract

This study explores supervised ML-DL based approaches for enhancing credit card fraud detection and improving financial risk management systems for EU banks. This research proposes an ensemble method based on majority voting (Hard Voting Classifier) of deep learning models to detect fraud transaction. Artificial Neural Network (ANN), Convolution Neural Network (CNN), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU) have been used as deep learning models. First, the most significant features that affect the type of transaction (fraud or not fraud) have been selected. After that, the ML-DL models were applied. The performance of the proposed approach is tested using a confusion matrix, recall, precision, F-measure and accuracy. The proposed method is tested using accurate data that consists of 540,099 transactions recorded in Kaggle repository dataset of two days based on European card holder for September, 2023. The result shows that the Random Forest (RF) model detected anomalies with 99.99% accuracy, F1-score with 1.00, and excellent recall with 99.99%. As a result, the machine learning model based on RF approach shows promise as a real-time anomaly detection method with high performance and low computational cost.