Adaptive Fraud Detection in Online Transactions Using Machine Learning Techniques

Financial institutions suffer substantial losses dueto fraudulent online transactions. Traditional fraud detectionmethods often fail to identify evolving fraud patterns due todataset imbalance and high false negative rates. This paperproposes an adaptive fraud detection system integrating a valueat-risk (VaR) metric with machine learning techniques. Thesystem uses historical simulation to estimate potential fraudrelated losses and employs K-Nearest Neighbors (KNN) to classifyfraudulent transactions. The proposed approach enhances frauddetection accuracy while minimizing false negatives, providing aneffective fraud prevention framework for financial institutions.