AI-Driven Fraud Detection Models in Digital Payment Systems

Abstract

Digital payment systems have evolved rapidly, changing the way we carry out financial transactions by allowing consumers to make fast, secure and easily accessible cashless payments. At the same time, the immense increase in volume of electronic transactions created greater exposure to fraud and cyber risk. Traditional rule-based fraud detection techniques are becoming less effective as they don’t adapt to new types of fraud or the high volume of real-time transactions. This paper presents a systematic review of the literature and a comparative analytical synthesis of artificial intelligence (AI) and machine learning (ML) based fraud detection methods in digital payment systems. The results from the literature review indicate that AI based methods have been consistently better than all other approaches at detecting fraud, especially in situations where data is heavily imbalanced and transaction behavior is dynamic. Ensemble methods - especially random forest and gradient boosting - consistently outperform rule-based systems for detecting fraud in imbalanced transactional environments. The study results also indicate that AI will improve operational efficiencies, reduce financial losses, and build consumer trust through greater detection of fraudulent activity; and provide a uniform analytical framework for preventing fraud in today’s digital financial ecosystems.