Nonlinear Modeling and Forecasting of Occupational Tax Revenues in Iranian Economy (An Application of Artificial Neural Networks and Its Comparison with Linear Regression and Time Series Models)

In this paper, in order to have a more a accurate forecast, three kinds of models: linear regression, time series and artificial neural network are used. Using Lyapunov exponent test, the structure of time series based on linearity, nonlinearity and stochastic process is studied. The results show a weak chaos in the system and declare possibility of applying nonlinear modeling for short-run forecasts.Using artificial neural network, linear regression and time series methods, occupational income tax are forecasted for period 1379-1383 and their performance is compared with each other. The results indicate better performance of neural network model.