A research on the use of data mining methods in determining financial statement frauds

In this study, it is aimed to identify fraudulent financial reporting practices of companies whose stocks are traded in Borsa Istanbul Textile, Apparel and Leather sector through financial ratios by using methods based on data mining, and accordingly, to reveal the power of data mining methods in detecting fraud. In accordance with the purpose of the study, first of all, the financial statements bearing the risk of fraud were determined by means of the independent audit reports of the companies included in the study and the weekly bulletins of the Capital Markets Board. In this way, companies are divided into two groups as companies that apply for financial statement fraud and companies that do not resort to financial statement fraud. Then, the financial statements bearing the risk of fraud and not carrying the risk of fraud were compared according to some selected financial ratios. Finally, the ability of some methods based on data mining to detect financial statements that are considered to have the risk of fraud and financial statements that are deemed to have no risk of fraud has been tested. According to the results of the research, Deep Learning Algorithm and J48 Algorithm, which are among the methods based on data mining, are the most successful methods at 84.25% in detecting the risk of fraud, and these methods are 54.29% successful in detecting the risk of fraud and in detecting financial statements that accept that there is no risk of fraud. It was concluded that it was successful with a rate of 95.65%. Keywords: Mistake, Fraud, Financial Statement, Fraudulent Financial Reporting, Data Mining