Detection of Mule Accounts and Fraudsters in UPI Transactions Using AI and Machine Learning Techniques

The rapid expansion of the Unified Payments Interface (UPI) ecosystem in India has been accompanied by a significant rise in digital financial fraud, particularly through the use of mule accounts - bank accounts used by fraudsters to launder illicitly obtained funds. Conventional rule-based fraud detection systems are increasingly inadequate against the sophistication and volume of modern UPI fraud. This paper proposes a comprehensive multi-model machine learning framework that integrates Gradient Boosted Decision Trees (GBDT), Graph Neural Networks (GNN), and Long Short-Term Memory (LSTM) networks to detect mule accounts and fraudulent actors in UPI transaction data. The proposed system analyses transaction behaviour, network topology of fund flows, and temporal transaction patterns to accurately identify suspicious accounts. Experimental results demonstrate a detection accuracy of 94.3%, precision of 92.7%, recall of 91.8%, and an F1-score of 0.923, outperforming existing single-model approaches. The framework offers a scalable, interpretable, and real-time deployable solution for banks, payment service providers, and financial regulators