Title
Identifying Fraudulent Financial Reports: Verification Between the M-Score Model and the Auditor’s Opinion
Authors
Abstract
Purpose: This study examines and forecasts financial reporting fraud in listed enterprises using the M-score model and auditor opinions based on the fraud triangle model.
Methodology: Research data was collected from non-financial enterprises listed on the HSX and HNX exchanges from 2018-2022. This study uses today’s popular machine learning methods to evaluate the performance of models to have a basis for recommendations (machine learning methods such as ANN, KNN, Decision Tree, and Random Forest) and gradient boosting algorithms (XGBoost and LightGBM). These methods help make decisions more accurately and help financial managers identify fraudulent financial reporting of companies early. This is consistent with requirements in management science and decision sciences.
Findings: The ANN model for the M-Score achieved the highest accuracy (97.9%) and F1-score (0.979). In comparison, the Decision Tree model was most effective for auditor opinions with an accuracy of 82.1% and an F1-score of 0.831. Additionally, the XGBoost algorithm consistently delivered strong results across both models, with an F1-score of 0.984 for M-Score and 0.942 for auditor opinions.
Originality/Value: In this article, this study relies on the fraud triangle theory, briefly finding the elements of the three factors from the fraud triangle model, combined with the auditor’s opinion on all financial statements. From there, predict whether a company has fraudulent financial statements or not. This way, this study combines the financial statement fraud theory with reality based on auditors’ comments. In addition, this study also compares the traditional forecasting method, M-score, to evaluate the performance of forecasting models.
Implications: The auditor opinion model holds practical value, integrating qualitative and quantitative insights for early fraud detection.
Limitations: Further empirical research is required to select indicators representing identifying signs in the fraud triangle model. The model based on auditors’ opinions holds significant reference value as it integrates qualitative and quantitative aspects, thereby combining theory with practical application.
Keywords
Financial statement fraud, machine learning methods, Vietnamese listed companies, M-Score, Auditor opinion
Classification-JEL
C11, C23, C53, E37
Pages
23-45
How to Cite
Phong, N. A., Tam, P. H., & Thanh, N. P. (2024). Identifying Fraudulent Financial Reports: Verification Between the M-Score Model and the Auditor’s Opinion. Advances in Decision Sciences, 28(4), 23-45.