Automated Z-Score method for fraud detection in rural credit banks

Abstract

This research proposes a new transaction monitoring framework using AIS, integrating Z-Score and Interquartile Range methods into a real-time system for fraud detection and early warning in Rural Credit Banks. The previous studies on fraud detection have been mainly oriented towards analytical models and machine learning approaches. Very little attention has been paid to integrating lightweight statistical anomaly detection approaches in operational Accounting Information Systems. Unlike traditional AIS-based monitoring systems that rely on periodic audits and manual supervision, the proposed framework enables continuous statistical anomaly detection in a real-time AIS environment based on computationally efficient methods. The main contribution of this work is not the development of new statistical techniques, but the operational integration of statistical anomaly detection methods into an AIS-based monitoring environment. The framework was conceptually designed using an Agile Software Development approach to classify transactions as normal or suspicious based on Z-Score and IQR calculations. A preliminary empirical validation was conducted on a synthetic banking transaction dataset with 852 transactions. The framework also achieved detection accuracy of 98.00%, precision of 92.38%, recall of 100%, and F1-score of 96.04%. The findings indicate that the incorporation of statistical anomaly detection methodologies within AIS environments can improve internal control, boost the efficiency of transaction monitoring, and support the prevention of fraud in Rural Credit Banks and similar financial entities.