Application of Data Mining Techniques in the Detection of Financial Statement Fraud

Financial statement fraud is one of three forms of fraud (financial statement fraud, corruption and asset misappropriation) that causes the greatest losses for companies. The increase in the number of financial statement frauds over the past 20 years (Enron, WorldCom, Parlamat, Tesco, Wirecard…) together with the development of data mining techniques, which are a prerequisite for the early detection of fraud, has led to a greater number of studies in this field. To date, research that has been carried out on the detection of financial statement fraud has not given rise to an unambiguous model; the various studies have suggested the use of different financial analysis indicators and have used different data mining methods, which has led to various levels of success in the detection of financial statement fraud. The reason for this is the fact that these studies have not taken into account changes in the business environment, which have a direct influence on the structure of financial statements, and, by this means, also on the financial analysis indicators. This study shows that focusing on shorter time periods and on companies in identical markets can lead to a better quality model for detecting financial statement fraud