The evaluation accuracy of BBO and ICDE as Linear- evolutionary Algorithms and SVR and CART as Non-linear Algorithms to earnings management prediction

Environmental uncertainties associated with new business models causes the complexity of the decisions. In such conditions, It can not be analyzed data with traditional means because many of affecting variables was unknown and them relationships are nonlinear and complex too. Therefore In this study, the accuracy of Artificial Intelligence Algorithms to predict earnings management is investigated. To predict the earnings management, the Modified Jones Model we have used and Biogeography-Based Optimization (BBO) and Improved (μλ) Constrainted Differential Evolution (ICDE) as Linear- evolutionary Algorithms and Support Vector Regression (SVR) and Classification and Regression Trees (CART) as Non-linear Algorithms were compared to investigate the power prediction of earnings management. The statistical population of the research, is manufacturing companies listed in Tehran Stock Exchange and 89 companies during 2007 to 2014 were studied. The results indicate that in general, Nonlinear Algorithm is more accurate than the Linear Algorithm and Support Vector Regression Algorithm is better than other algorithms to predict the earnings management. Also, Linear Algorithms showed roughly similar results to predict earnings management.