Artificial Intelligence-Powered Credit Card Fraud Detection: Feature Engineering and Machine Learning Approach’s

Abstract

The payment experience has been completely transformed by the use of cashless payment options including credit card purchases and web transactions, but it has also led to more sophisticated financial fraud, posing a significant challenge to payment system security. Accurately detecting fraudulent transactions while reducing false positives is a need for credit card fraud detection systems. use of a Convolutional Neural Network (CNN) model to detect fraudulent transactions is examined in this study using the Kaggle Credit Card Fraud Detection dataset. The CNN model performed quite well, with an F1 score of 79.52%, accuracy of 99.93%, precision of 80.8%, and recall of 78.29%. With a balanced trade-off between accuracy and recall, these findings demonstrate the model's capacity to detect fraud and manage unbalanced datasets. Further evidence of CNN's higher performance comes from comparison with
other models, including k-Nearest Neighbours (k-NN) with Random Forest. This study demonstrates how advanced deep learning methods may be applied to effectively detect credit card fraud. Future research can explore hybrid models, advanced deep learning techniques, and domain-specific feature engineering to enhance model robustness and adapt to evolving fraud patterns.