Adaptive Hybrid Machine Learning Framework for Real-Time Detection and Mitigation of High-Value Financial Fraud in Digital Payment Ecosystems and Platforms
Main Article Content
Abstract
The rapid growth of digital payment systems in India has enhanced transaction efficiency and financial inclusion, but it has also led to a significant increase in digital payment fraud. This study analyzes trends in digital payment fraud in India from 2018 to 2023 using secondary data, focusing on changes in fraud frequency and associated financial losses. The analysis reveals a steady rise in fraud cases, increasing from 145,000 in 2018 to 410,000 in 2023, along with a disproportionate growth in financial losses from ₹710 crore to ₹3,600 crore. The findings indicate that fraud incidents are increasingly concentrated in high-value transactions, amplifying economic damage. The study further highlights the limitations of traditional rule-based and standalone machine learning models, which exhibit low recall and poor adaptability to evolving fraud patterns. In contrast, hybrid machine learning models integrating ensemble methods with reinforcement learning demonstrate superior accuracy, recall, and real-time detection capability. The study emphasizes the need for adaptive, cost-sensitive, and risk-aware fraud detection frameworks in modern digital payment ecosystems.