AI-Driven Marketing Framework for E-Commerce Transactions: Fraud Reduction and Customer Retention
DOI:
https://doi.org/10.64252/mb2zw275Keywords:
Unified Payments Interface (UPI), fraud detection, customer retention, reinforcement learning (RL), revenue optimizationAbstract
Background: India’s Unified Payments Interface (UPI) processed 13,303.99 million trans- actions worth 2,839.52 billion USD in April 2024, facing challenges in fraud, transaction declines, and competition.
Methods: Using 45 million transactions (2021–2024), the framework integrates multi- level modeling, K-means clustering, Long Short-Term Memory (LSTM) neural networks, and reinforcement learning (RL) to optimize user targeting, segmentation, fraud detection, customer relationship management (CRM), and time-limited price promotions.
Results: It achieved a 58% fraud reduction, 19% decline reduction, 16% retention increase, 15-point Net Promoter Score (NPS) rise, and 128.76 million USD in revenue. Urban peer- to-merchant (P2M) users drove 60% of volume with a 2.80 USD customer lifetime value (CLV).
Conclusions: The framework, the first multilevel approach for UPI, offers Payment Ser- vice Providers (PSPs) actionable strategies for loyalty and inclusion, with future work exploring blockchain and lightweight AI models.
