A Bayesian Network-Based Model for Fraud Risk Assessment

Global practice shows that almost every second company is exposed to financial fraud, while the process of its detection is labor-intensive and often low-efficient. This paper proposes a model for assessing the risk of fraud committed by business entities based on a Bayesian network. The model adopts a modern methodology for classifying fraud risk factors, referred to as the Fraud pentagon. The evaluation within the model incorporates financial statements, accounting data, and expert assessments regarding the internal controls. The effectiveness of classifying companies as fraudulent and bona fide using the model has been experimentally tested. It has been found that the risk-oriented approach underlying the model makes it possible to substantially reduce labor inputs for audit while maintaining high credibility of the results.