A framework combining quantitative analytical methods to detect anomalies in financial statements of organisations
Date
2024-12
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch University
Abstract
This thesis develops the Annual Financial Statement Risk Alert Signal (AFS RAS) framework to detect financial statement fraud and anomalies, integrating the Altman Z–Score, Beneish M–Score, and Benford’s Law. The framework aims to identify fraudulent activities within organisations by assessing financial health and detecting anomalies in reported data. The thesis evaluates a selection of major corporations across various industries in the United States of America, highlighting the effectiveness of each method in detecting potential fraud. The Altman Z–Score provides insights into financial distress, the Beneish M–Score identifies possible earnings manipulation, and Benford’s Law detects anomalies in numerical data. The results indicate that the framework achieves a True Positive Rate (TPR) of 80%, successfully identifying 20 out of 25 fraudulent financial statements, with a False Positive Rate (FPR) of 20%. The analysis shows that revenue manipulation is the most common type of fraud, followed by earnings manipulation and accounting issues. While the AFS RAS framework demonstrates strong potential in fraud detection, further refinement is needed to reduce false positives and enhance accuracy. This thesis highlights the importance of comprehensive financial analysis in improving fraud prevention strategies and safeguarding organisational integrity. Future research should focus on integrating additional machine learning techniques and expanding the dataset to include broader data sources and international contexts to enhance the framework’s efficacy further.