Comparison of Earnings Management Prediction Using Neural Networks Model and Modified Linear Jones Model

Purpose of this research is studying and comparing neural network prediction and Jones linear model in order to predict earnings management of firms accepted in Tehran Stock Exchange. Research statistical sample includes all firms accepted in Tehran stock exchange during 2007–2011 and sample size is 160 concerning screening method and after eliminating outlier observations. Results of data analysis showed that prediction power of earnings management has significant difference with using neural network and Jones modified linear model. Jones modified linear determination factor is 51.1% which indicates that this model can predict 51.1% of firms’ earnings management changes. On the other hand, results obtained by implementing neural network show that this model can predict 92.8% of variations in earnings management.