Enhanced Credit Risk Prediction Using Ensemble Learning with Data Resampling Techniques

Credit cards are now the most popular mode of payment for both offline and online purchases due to new developments in electronic commerce. Consequently, fraudulent credit card transactions have surged, causing substantial financial losses to businesses and individuals annually. This research addresses the challenges of credit card fraud detection including severely imbalanced datasets, evolving fraud techniques, and high false-positive rates. We propose an Enhanced Credit Risk Prediction system using a Gradient Boosting Classifier combined with SMOTE-based data resampling to overcome class imbalance. Comprehensive empirical analysis was conducted using the European Card Benchmark dataset. The evaluation demonstrates optimized results: Accuracy of 99.9%, F1-Score of 85.71%, Precision of 93%, and AUC of 98%, outperforming existing machine learning and deep learning approaches including ANN and CNN. The proposed system provides a practical, deployable solution for real-world credit card fraud prevention.