FRAUD DETECTION IN CREDIT CARD TRANSACTIONS USING MACHINE LEARNING: A COMPARATIVE ANALYSIS

Abstract

The rapid increase in digital financial transactions has expanded the risk of fraudulent activities, which escalated the need for more effective fraud detection methods. This research paper presents a comprehensive approach to fraud detection in credit card transactions, utilizing an end-to-end methodology based on machine learning techniques. This study summarizes thorough exploratory data analysis, data preprocessing, model training with various class balancing techniques, and hyperparameter optimization. Experiments on a large credit card dataset (284,807 transactions, 31 features) have demonstrated that the best model - XGBoost without sampling - achieved an Area Under the Curve of 96.82% and an average precision of 88%. Since the framework is trained using a data-driven approach, it can dynamically adapt to emerging fraud patterns, ensuring high accuracy in identifying suspicious activities. Additionally, the capability of the model to handle large volumes of data in real-time makes it well-suited for financial institutions managing high transaction loads. One key advantage of this architecture is its ability to achieve significantly higher throughput while maintaining low latency compared to traditional methods.