Recent scandals and corporate failures have shaken the confidence of investors as accounts which were purported to reflect a “true and fair view” of businesses have been misleading. In most of the reported cases management have defrauded the company and covered it up by manipulating the financial statements of the company to reflect what management wanted the public to see. Statistics and machine learning based technologies have been shown to be an effective way to deter and defect fraud. Therefore, this study deals with the identification and selection of financial and non-financial variables related to fraudulent financial statements and investigates the effectiveness of Data Mining techniques in detecting firms that issue fraudulent financial statements. To achieve this aim 40 financial and non-financial variables derived from financial statements of listed companies in Tehran Stock Exchange (TSE). We consider Data Mining based financial fraud detection techniques such as Artificial Neural Networks (ANN), Bayesian Networks (BN) and Random Forest (RF) in order to identify fraud. The results suggest that Data Mining techniques have a good ability to detect financial statements fraud. The three models are compared in terms of their performance. The results of Wilcoxon signed-rank test indicates that The Random Forest algorithm and Bayesian Networks outperforms the other model. Artificial Neural Networks also achieve good performance.