Unveiling the Optimal Approach for Credit Card Fraud Detection: A Thorough Analysis of Deep Learning and Machine Learning Methods

This study compared machine learning (ML) and deep learning (DL) techniques for credit card fraud detection.We evaluated 16 combinations of ML algorithms and cross-validation methods across diverse datasets.The Random Forest classifier with repeated K-fold cross-validation achieved the highest accuracy 99.0% and F1 score 99.1% among all models.The top performing deep learning model, the Artificial Neural Network (ANN), achieved an accuracy of 91.3% and F1 score of 91.1%, while a hybrid model combining these approaches reached 98.9% accuracy and F1 score.The Random Forest Classifier continued to be the best option.Our findings suggest the Random Forest classifier with repeated K-fold crossvalidation, tested against a 21 combinations of other machine learning models, deep learning models, and a hybrid model as the most reliable method for credit card fraud detection in balanced datasets, offering valuable insights for enhancing security precautions and financial system defense against various banking sector frauds.