New Approach to Predicting and Detecting Financial Statement Fraud, Using the Bee Colony

Abstract

Objective: Considering complex financial plans to conceal fraud in financial statements, the development of fraud detection methods can be regarded as solution for this problem. The present study uses the bee algorithm to develop methods for fraud detection in financial statements. Method: Three methods of bee algorithm, genetic algorithm and logistic regression have been used to study the subject. The statistical sample consists of 120 companies accepted in the Tehran Stock Exchange (60 companies are suspected of fraud and 60 ones are not suspected) for the period 1396-1385. The companies were suspected of fraud, based on 1) revised audit opinion after unacceptable expression, 2) existence of significant annual revisions, and revised financial statements for inventories and other assets; 3) existence of tax disputes with the tax area, according to notes on income tax filing, general tax filings and conditioned clauses in audit reports. Following the use of cross-entropy, 16 financial ratios were introduced as the potential predictors of fraudulent financial reporting. Result: The results showed that the bee algorithm method with prediction accuracy of 82.5% has better performance in identifying suspicious companies in fraudulent financial statements than the other two methods. Conclusion: The results of the research indicate that the proposed method of this study compared to other methods has higher rate of prediction accuracy, lower error rate and relatively good speed.

Keywords

• detecting of fraud in financial statements
• bee algorithm
• mutual entropy
• financial ratios