Credit Card Fraud Detection Utilizing Machine Learning, Deep Learning, and Bayesian Networks

Detecting credit card fraud is a critical challenge in the modern financial landscape, where robust and efficient solutions are essential to mitigate losses and safeguard consumers. This research evaluates and contrasts three primary approaches to this issue: Machine Learning (ML), Deep Learning (DL), and Bayesian Networks (BN). Using publicly available datasets, several ML techniques, including Logistic Regression, Support Vector Machines (SVM), Random Forests, Decision Trees, and XGBoost, were assessed alongside DL models such as Multi-Layer Perceptron (MLP) and Deep Neural Networks (DNNs). Additionally, a Bayesian Network was employed to analyze causal relationships between fraud-related features, offering an interpretable probabilistic framework. While ML and DL approaches achieved high performance regarding accuracy, precision, and recall, the Bayesian Network distinguished itself by providing enhanced interpretability and reliability. This study highlights the complementary strengths of these methods and the potential for their integration to develop adaptive and efficient fraud detection systems