Fraud Detection in Financial Transactions Using Machine Learning with Oversampling Techniques: A Case Study of a Moroccan Bank

Abstract

Fraud detection in financial transactions is essential for ensuring the security and integrity of financial systems. This research applies machine learning models to detect fraudulent transactions in a highly imbalanced dataset from a Moroccan bank, where fraudulent transactions account for only 1.8% of the total records. To address this imbalance, oversampling techniques, specifically SMOTE (Synthetic Minority Over-sampling Technique), were applied. The study implements various machine learning models such as Random Forest, Support Vector Machines (SVM), Logistic Regression, and Gradient Boosting, which have been widely discussed in recent literature. The performance of these models is evaluated using precision, recall, F1-score, accuracy, and ROC-AUC. Gradient Boosting, combined with SMOTE, emerged as the most effective model in detecting fraudulent transactions. This paper includes detailed definitions of machine learning techniques, a review of related literature, and a comprehensive comparison of the models used.