Data-driven auditing: A predictive modeling approach to fraud detection and classification

This article develops and empirically tests a predictive model for audit of fraud detection with practical applications for audit operations. By analyzing real-life accounting data, the proposed model can identify anomalous transactions and directly focus on exceptions for further investigation in real time, thus offering a significant reduction in manual intervention and processing time in audit operations. Our approach is a highly desirable supplement to the existing rule-based models, given the growing use of information technology for analytics in auditing. The proposed approach is based on classification. Following the tenets of the principal agency theory, we discuss how our approach can help to reduce monitoring and contracting costs, disincentivize fraud, improve auditor efficiency and independence, and increase audit quality. We contribute to the current literature by discussing the implications of data-driven audit on the moderating role of auditors in principal-agent relationships and providing practical insights into the operational aspects of financial reporting and auditing, modeling of fraud-detection classification models, and benefits, barriers, and enablers of implementing data driven audit in companies.