Top level management is usually found responsible for fraudulent reporting of financial statements to fulfill the objective of artificially improving the financial performance and results of the company. Several data mining algorithms have been implemented for successful identification of fraudulent financial reporting. This paper explores the four commonly used data mining techniques, viz., Neural Network, Decision Trees, Genetic Algorithms and Bayesian Belief Networks for detection of financial statement fraud. This study investigates the effectiveness of these four techniques in identifying fraudulent financial statements. In addition, the techniques were compared in terms of their performances based on eight varying parameters. Neural network appeared as the most extensively used technique for detection and identification of financial statement fraud.

Financial statement fraud (FSF) costs billions of dollars every year to the world’s economy. The top executives of organizations such as Satyam Computer Services Limited (India’s 4th biggest software services exporter), Enron Corporation and WorldCom, were accused of manipulating the books of accounts. The fall of Satyam Computer Services Limited, caused by alleged financial statement fraud, is the biggest bankruptcy in the Indian history, and that of WorldCom in the US. 78 bn was the cost of fraud confessed to by B Ramalinga Raju, the then chairman Satyam, by falsifying financial statements (The Hindustan Times, 2009).

Generally, the perpetrators of fraud come from within the organizations. In more than 40% cases, fraud has been perpetrated by top executives, including board members, directors, etc. (Figure 1) (KPMG Report, 2008).

Systems Management Journal, Unsolicited Bulk E-Mails, International Spamming, Pornographic Spam e-mails, Hindi Language in e-mails, Internet Slang, Slang Hindi Words, Tokenization, English Slang Words, Online Communication, Hindi Words Vector, Non-spam e-mails.