Explaining the model of earning management measurement using an intelligent hybrid method of neural networks and meta heuristic algorithms (Genetic and particle swarm optimization)

Subject Areas : Financial engineering

Eghbal Ghaderi ¹ , piman amini ² , Iraj Noravesh ³ , Ata Mohammadi Moqrny ⁴

1 - Departmental accounting،, Azad University of Sanandaj, Sanandaj,Iran
2 - Department Accounting The University of Kordestan, Sanandaj,Iran
3 - Prof., Department of Accounting, Sanandaj Branch, Islamic Azad University, Sanandaj, Iran
4 - Assistant Prof., Department of Accounting, Sanandaj Branch, Islamic Azad University, Sanandaj, Iran

Received: 2018-01-26 Accepted : 2018-05-21 Published : 2018-09-23

Keywords: Genetic Algorithm, Particle Swarm Optimization, Earnings Management, Neural network,

Abstract :

Undrestanding the earning management for the users of accounting information due to performance evaluation, profitability forecast and detrmining the value of the company is very important.The purpose of this research is to estimate the a model for earning management using neural network model and then the use of Genetic Algorithm, and Particle Swarm Optimization to find a better combination of input data, so that it can optimize the initial model. For this purpose, 28 effective variables were used in the from of four groups (Financial, managerial, corporative and audit) during the years 2010 to 2016 in the companies admitted to the Tehran stock Exchange. The results showed that application of this algorithm has increased the efficiency of the model.Also, the evaluation of the performance of neural network patterns suggests the absolute superiority of this pattern compared to the time linear method (LR).Combined method (A-PSO) and (A-GA)by identifying four optimal variables respectively precision forecast, shareholding of major shareholders, company size and the ratio of the quality of earning management are carefully predicted respectively (%95/59) and (%94/75). In addition to the above mentioned intelligent methods, by improving correlation coefficient and error squares mean criterion compared to linear methods (LR) and neural network method (ANN) in predicting group results, management and corporate features are more efficient.