Following the development and establishment of public institutions, the separation of ownership and management became necessary. In result, agents took the control of such institutions and practically the management separated from ownership. The managers have the role of supervision and accountability against the resources and preparing financial reports as well. The conflict of interests between managers and owners increases the risk of unreliable information. Earnings as the most important information which available for users, can be manipulated by managers' various motivations. Thus , the main aim of this research is to predict earning management by decision trees in order to help investors.
 In this research , the independent variables consist of ownership percent of institutional stockholders , debt rate, firm size , income  tax , sale variability,  earnings variability , cash flow from operating activities , interest quality rate, total assets flow , sale return , return on investment and stockholders' rate of return and the discretionary accruals are dependent valuable as well. The results of current research show that the highest prediction level for decision trees is 74.7 %.