The Detection of Financial Statements Fraud According To Audit Report of Financial Statments

This paper aims the detection of financial statements fraud according to audit report of financial statments. The initial research data were collected from a statistical sample consisting of 164 companies, listed in the Tehran Stock Exchange from 2014 to 2017 and selected through the systematic sampling method. The statistical sample was divided into two separate groups, i.e. fraudulent (1) and non-fraudulent (0) companies. The independent fraud-related variables included 41 financial and nonfinancial variables, selected through theoretical foundations and the research background. The data of variables, collected through the desk method, were finally analyzed through the top five techniques of machine learning, including; the Bayesian networks, the decision tree, artificial neural networks, support vector machine, and combined method. According to the results, all of these techniques were highly capable of fraud detection in financial statements. Moreover, the proposed combined technique outperformed the other techniques in evaluation accuracy and power with an estimation rate of 96.2%.