ABSTRACT

digital payment systems have significantly increased the risk of online transaction fraud, posing serious challenges to financial security. This paper presents a machine learning–based fraud detection system developed using the MERN (MongoDB, Express.js, React.js, and Node.js) stack. Unlike conventional approaches that rely on Python-based frameworks, the proposed system implements a Gaussian Naive Bayes classifier in a JavaScript environment to enable seamless integration with web applications. A synthetic dataset is generated to simulate real-world transaction scenarios using key features such as transaction amount, device type, geographical location, and inter-transaction time gaps. The model is trained to classify transactions as legitimate or fraudulent while also providing a probabilistic fraud score to enhance decision-making. The system architecture incorporates both user and administrative modules, enabling secure transaction processing, real-time fraud detection, and monitoring through an interactive dashboard. Experimental evaluation demonstrates that the proposed model achieves high detection accuracy and effectively identifies suspicious transaction patterns. The system enhances the reliability and security of online financial operations while reducing the risk of fraud in digital payment environments.

Key Words— Fraud Detection, MERN Stack, Naive Bayes, Machine Learning, Online Transactions.