Related Experiment Videos
Enhancing credit card fraud detection with a hybrid approach using machine and deep learning
Nagwa Gamal1, Eman M G Younis2, Waleed M Makram3,4
1Faculty of Computers and Information, Minia University, Minia, Egypt.
Scientific Reports
|March 28, 2026
Summary
This study enhances credit card fraud detection using advanced machine learning (ML) and deep learning (DL) models. Novel stacking ensemble approaches significantly improved accuracy and performance, outperforming traditional methods.
Area of Science:
- Computer Science
- Artificial Intelligence
- Data Science
Background:
- Traditional rule-based credit card fraud detection systems are increasingly ineffective against sophisticated fraud tactics.
- High-value financial losses occur annually due to credit card fraud, impacting institutions and consumers.
- The imbalanced nature of transaction data, with rare fraudulent instances, poses a significant challenge for detection models.
Purpose of the Study:
- To enhance the accuracy and efficiency of credit card fraud detection systems.
- To address the challenge of imbalanced datasets in fraud detection using advanced sampling techniques.
- To evaluate and compare the performance of various machine learning and deep learning models, including ensemble methods.
Main Methods:
- Applied data balancing techniques like Synthetic Minority Over-sampling Technique (SMOTE) and SMOTE-Edited Nearest Neighbors (SMOTE-ENN).
- Evaluated diverse machine learning (ML) and deep learning (DL) models, including Extra Trees (ET), Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and eXtreme Gradient Boosting (XGBoost).
- Developed and assessed two novel stacking ensemble approaches, integrating multiple base models with meta-learners, and utilized SHAP and LIME for interpretability.
Main Results:
- The proposed stacking ensemble models achieved exceptional performance metrics, including accuracy, precision, recall, F1-score, and Area Under the Curve (AUC) approaching 1.0.
- Ensemble learning techniques (CatBoost, XGBoost) and deep learning models (FFNN, ANN, MLP) demonstrated superior performance over traditional methods.
- Comprehensive error analysis confirmed the robustness and transparency of the proposed fraud detection approaches.
Conclusions:
- Combining machine learning, deep learning, and data balancing techniques is crucial for effective credit card fraud detection.
- The developed stacking ensemble methods represent significant advancements in improving fraud detection accuracy and reliability.
- Interpretable AI techniques (SHAP, LIME) enhance the understanding and trustworthiness of complex fraud detection models.