A Hybrid Random Forest and Neural Network Model for Enhanced Credit Card Fraud Detection in E-Commerce

Credit card fraud continues to be a significant issue in e-commerce, resulting in financial losses and security risks. Conventional fraud detection techniques have challenges with imbalanced datasets and the adaptation of fraudulent strategies, requiring a more advanced methodology. This study presents a hybrid model combining Random Forest and Neural Network techniques that use feature selection, deep learning, and tuned hyperparameters to enhance fraud detection efficiency. Experimental findings indicate an accuracy of 99.2 %, precision of 98.8%, recall of 97.5%, and an AUC-ROC score of 99.3 %, greater than individual machine learning models. The method minimizes false positives and improves real-time fraud detection. This strategy provides enhanced generalization, resilience, and efficiency in real deployment compared to earlier methods. Future study may examine adaptive fraud detection techniques utilizing real-time streaming data and reinforcement learning to improve scalability.