Comparative Analysis of Machine Learning Algorithms for UPI Fraud Detection

In India, digital payment systems have revolutionised financial transactions; one of the most popular platforms is Unified Payments Interface (UPI). However, the quick expansion of UPI transactions has also resulted in a notable increase in fraudulent activity, which puts users and financial institutions at grave risk. Large-scale transaction volumes and changing fraud patterns are frequently too much for traditional rule-based fraud detection systems to handle. This paper suggests a machine learning-based fraud detection system for spotting dubious UPI transactions as a solution to this problem. Numerous classification techniques, such as Logistic Regression, Decision Tree, Support Vector Machine (SVM), K-Nearest Neighbours (KNN), Naïve Bayes, XGBoost, and a Voting Classifier ensemble model, are evaluated by the suggested system. Prior to model training, the dataset is preprocessed utilising feature scaling, encoding, and data cleaning methods. Metrics for accuracy, precision, and recall are used to evaluate performance. Strong predictive capacity across models is demonstrated by experimental data, suggesting that machine learning techniques may successfully differentiate between fraudulent and authentic transactions. To improve financial security and lower the risk of fraud, the created method can be included into real-time payment platforms.