An Innovative Hybrid Model Combining CNN and Random Forest for Financial Transaction Analysis

Financial fraud is a growing concern in digital transactions, with traditional fraud detection systems struggling to keep up with evolving fraudulent tactics. This study proposes a hybrid approach combining Convolutional Neural Networks (CNN) and Random Forest (RF) to improve real-time fraud detection accuracy. The CNN model extracts key features from financial transaction images, such as check forgeries and credit card receipts. At the same time, RF classifies the transactions as fraudulent or legitimate based on learned patterns. The proposed model was evaluated using the Kaggle Credit Card Fraud Dataset. The results showed that the CNN-RF hybrid model achieved an accuracy of 97 %, outperforming traditional Random Forest and CNN models, which had accuracies of 92 % and 93 %, respectively. The hybrid approach also demonstrated a high AUC score of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{0. 9 5}$</tex>, confirming its ability to differentiate fraudulent from non-fraudulent transactions. Further analysis indicated that fraudulent transactions tend to involve slightly higher amounts, making them distinct from legitimate ones. This research highlights the advantages of integrating deep learning with ensemble machine learning for fraud detection. The proposed method offers improved detection capabilities, reducing false positives and negatives and providing a more effective solution for financial institutions to combat fraud in real-time transactions.