Enhanced UPI Fraud Detection Using Machine L...

Enhanced UPI Fraud Detection Using Machine Learning: A Proactive Approach to Secure Digital Transactions

Author Name : Meena V, Harshitha M, Kavitha S

ABSTRACT With the increasing adoption of digital payments, detecting fraudulent transactions in UPI (Unified Payments Interface) systems has become a critical challenge. This paper presents a machine learning-based fraud detection system designed to identify and prevent fraudulent transactions in real time. Feature engineering techniques are applied to preprocess transaction data, and fraud classification is initially performed using the Random Forest algorithm. Anomalous transactions are further detected using Isolation Forest and One-Class SVM. Real-time fraud detection is achieved through Streaming Data Processing with Logistic Regression, ensuring immediate identification of suspicious activities. The system's performance is evaluated using Cross Validation, with effectiveness measured through the ROC Curve. Finally, fraud patterns are visualized using Heat map Visualization, and t-SNE is utilized for dimensionality reduction and clustering. This comprehensive approach improves fraud detection accuracy and provides valuable insights into fraudulent transaction patterns, enhancing the security and reliability of UPI-based transactions.