Machine learning model for real-time digital banking fraud detection

Abstract

Fraudsters use financial fraud to deceive for the purpose of financial gains. It has now become an international threat for companies and organizations. In Kenya, almost all financial institutions, especially banks, and insurance firms, have been victimized by financial fraud in one way or another. Traditionally, manual verification, inspection, and other such methods have been employed to identify fraudulent activities, without integrity, high in cost, and time-consuming. Artificial Intelligence and Machine Learning now provide a more efficient means for intelligent detection and prevention of fraudulent transactions, through the analysis of large financial datasets. The rapid growth in digital banking in Kenya has fairly promoted financial inclusion, but it also accelerated fraudulent activities that threaten financial institutions and their customers. The old rule-based systems for detecting fraud can hardly keep up with new tricks from fraudsters. This study proposes an unsupervised machine learning model for real-time detection of fraudulent digital banking transactions in Kenya. Using unsupervised learning algorithms, the model will detect unique anomalies and suspicious activities that deviate from ordinary customer behavior within the Kenyan digital banking realm. The model will be trained and validated with a real-world dataset of Kenyan digital banking. Keywords: artificial intelligence; fraud; detection; fraudulent banking operations; machine learning; artificial intelligence.