Machine Learning-Driven UPI Fraud Detection

Financial transactions have been transformed by the emergence of digital payment platforms such as the Unified Payments Interface (UPI), which offer quick and easy services. But this ease of use has also raised the possibility of fraud, which presents serious problems for both consumers and financial institutions. This project offers a machine learning- based method for efficiently identifying fraudulent UPI transactions. To improve model accuracy, a wide range of data was gathered and preprocessed, including transaction frequency, amount, location, device information, and user behavior. Missing value handling, class imbalance correction, and feature selection using Lasso regression were among the preprocessing steps. To find important patterns connected to fraud, correlation and graphical analyses were performed. After several machine learning models were trained and assessed, the top-performing model was incorporated into an HTML, CSS, and Flask web application. The backend uses the trained model to instantly classify transactions as either legitimate or fraudulent after users safely enter transaction details. The findings show how machine learning can improve the security of UPI transactions by providing a scalable, effective, and user- friendly fraud detection solution that supports the development of digital financial systems.