Earning Management Prediction Applying Hybrid Multi-Layer Perceptron Neural Network and Meta-heuristic Algorithms

Accurately predicting of earning management to detect and identify manipulation of financial statements has always been one of the most fundamental Challenges ahead of financial reports users. Applying the Beneish model can be one of the most appropriate models for modeling in the field of earning management prediction. Beneish model 1999 has been developed for this purpose by emphasis on the variables of corporate governance system including audit committee structure, legal inspector and independent auditor, board of director's structure and corporate ownership structure requirements. The data of 81 companies listed on TSE during 2012-2018 has been analyzed by the hybrid method of multi-layer perceptron MLP neural network and metahuristic algorithms of biogeography based optimization BBO, Imperialist Competitive Algorithm ICA and water cycle algorithm WCA. The accuracy of the model by hybirid mothodes of MLP-BBO, MLP-ICA and MLP-WCA has been increased from 59.08 ، 59.96 and 59.79 percentages to 92.06 ، 89.24 and 79.72 percentage respectively. The results indicate the accuracy improvment of the proposed model in detecting earning-manipulator companies and also the higher efficiency of the hybrid method of MLP-WCA compared to the other methods in optimization of multilayer perceptron neural network.