Detection of Anomalies in Unified Payment Interface Using Machine Learning

Abstract

The rapid growth of the Unified Payments Interface (UPI) in India has transformed the digital payment landscape by enabling fast and user-friendly transactions.However, this rapid growth has also led to a significant rise in fraudulent activities, exposing the limitations of traditional rule-based fraud detection systems. To address this challenge, this project proposes a machine learning-based anomaly detection system designed specifically for UPI transactions. The system utilizes a stacking ensemble model that combines multiple classifiers—Support Vector Machine (SVM), Random Forest, K-Nearest Neighbors (KNN), and XGBoost—to enhance detection accuracy and reduce false positives. Feature selection is performed using the Chi-square test to isolate the most relevant transactional attributes. The solution is developed using Python in Visual Studio Code and deployed via a Flask web application with a MySQL database for data handling and real-time prediction. Experimental results demonstrate that the stacking model achieves high accuracy, outperforming individual models in terms of precision, recall, and F1-score. This project provides a scalable, real-time fraud detection framework capable of adapting to evolving financial threats in digital payment ecosystems. KEYWORDS: UPI, Fraud Detection, Machine Learning, Anomaly Detection, Stacking Ensemble Model, Chi-Square Feature Selection, Real-Time Prediction, Supervised Learning, Flask Web Application, Financial Security.