Hybrid Data Mining Technique for Credit Card Fraud Detection

The rising incidence of credit card fraud underscores the need for innovative strategies to enhance credit card fraud detection and prevention. Numerous approaches have been employed for credit card fraud detection; however, the field continues to seek methods that can adapt to the constantly evolving nature of fraud patterns. In this study, we develop a hybrid model by integrating machine learning algorithms for effective credit card fraud detection. Using a simulated credit card transaction dataset, the model is developed in two stages, the first stage finds a base algorithm for the proposed hybrid model. The second stage focus on developing the hybrid model by combining the base model (Light Gradient Boosting Machine) with each of the selected algorithms. The hybrid models, demonstrated superior performance compare to standalone algorithms. Also, the hybrid of LGBM and XGBoost model outperforms others combinations, having 98.3% accuracy, 98.88% Precision, 98.05% Recall, 98.46% F1-Score, 99.80% AUROC. This proposed hybrid model can enhance security and foster trust in financial institutions and businesses, and in turn contribute to a more stable and efficient financial ecosystem.