Fraud Detection Methods in Iran: A Comparative Study of CVX and LASSO Methods
Introduction
A recent study published in the Journal of Applied Accounting Research has shed light on the effectiveness of two data mining methods in detecting manager’s fraud risk in Iranian companies. The study, conducted by Alireza Rahrovi Dastjerdi, Daruosh Foroghi, and Gholam Hossain Kiani, analyzed the texts of board reports to identify words with the greatest power in explaining a company’s high fraud risk index.
Research Methods
The researchers used two methods - convex optimization (CVX) and least absolute shrinkage and selection operator (LASSO) regression - to identify words that were most strongly associated with a company’s likelihood of fraud. The study analyzed the texts of board reports from Iranian companies to detect manager’s high fraud risk index.
Results
The results showed that both methods were able to detect manager’s high fraud risk index with precision rates between 82.55% and 91.25%. However, the LASSO method was found to be significantly more precise than the CVX method.
Implications
The study highlights the importance of using data mining and text mining methods in detecting fraud, particularly in countries where financial statements may not always accurately reflect a company’s true financial situation. The authors suggest that regulatory bodies and independent auditors could use these methods to assess fraud risk for firms or other legal parties.
Limitations
The study notes some limitations, including the lack of access to an official list of firms suspected of fraud and the lack of Microsoft Word files for board reports. Nevertheless, the findings offer valuable insights into the capabilities of data mining and text mining in detecting manager’s fraud risk using board reports.
Significance
The study is the first of its kind in Iran and has significant implications for accounting and finance research in the country. As such, it could contribute to the development of more effective methods for detecting fraud and improving financial transparency in Iranian companies.
Key Takeaways:
- Both CVX and LASSO methods were able to detect manager’s high fraud risk index with precision rates between 82.55% and 91.25%.
- The LASSO method was found to be significantly more precise than the CVX method.
- Data mining and text mining methods can be used to detect fraud in Iranian companies.
- Regulatory bodies and independent auditors could use these methods to assess fraud risk for firms or other legal parties.