Enhancing Fraud Detection in Credit Card Transactions using XGBoost and SMOTE: A Comparative Study

The rise of online payment systems, particularly credit card transactions, has revolutionised commerce but also brought about a surge in fraudulent activities. Detecting these fraudulent transactions is crucial for financial institutions and businesses. Traditional machine learning methods exist for fraud prediction, but enhancing their accuracy is imperative. This study focuses on utilising ensemble techniques like XGBoost (eXtreme Gradient Boosting) in combination with the Synthetic Minority Oversampling Technique (SMOTE) to address class imbalance in fraud detection datasets. By conducting a comparative analysis against established machine learning methods, the research underscores the importance of ensemble learning and class imbalance handling in credit card fraud detection. Through experimentation, the XGBoost classifier coupled with SMOTE demonstrates promising results, achieving a precision score of 99.94% and an accuracy score of 99.9684% This approach offers a robust framework for effectively detecting fraudulent credit card transactions amidst the evolving landscape of financial technology.