Fraud Detection in Banking Transactions Using Machine Learning

Abstract

Consumer confidence has been steadily declining as a result of banking fraud, which has recently emerged as a major risk in the banking industry. Online transactions are becoming more sophisticated and numerous, and conventional methods of detecting fraud have not kept up with the increased number and velocity of these transactions. By examining user behaviour and transaction characteristics, this paper presents a machine learning-based approach to detect fraud in banking transactions. The system utilizes supervised learning algorithms, including Logistic Regression and Multinomial Naïve Bayes, to classify transactions as legitimate or fraudulent. A comprehensive administrative dashboard is developed to manage users, datasets, and classification outcomes efficiently. The architecture includes secure user authentication, dataset uploading, and performance analysis of classification models. The study offers a scalable and trustworthy solution for real-world banking applications, and experimental results show that the proposed system accurately identifies fraudulent activities. It also highlights the potential of machine learning to improve fraud detection systems, which could lead to future advancements using more complicated models and real-time processing frameworks.