Nonlinear Modeling and Forecasting Tax of Legal Entities

This paper deals with forecasting the tax revenues of legal entities in Iran. For this purpose, the structural natures of time series of tax revenues for Iranian legal entities are detected. Based on the separation among the resources (government and NGOs), the linearity, nonlinearity, chaotic, and random behaviors are diagnosed via the Lyapunov exponential analysis. Using Box- Jenkins and Neural Networks models with different numbers of input, output, hidden layers, learning algorithm, learning rate and etc., the performance of each model are evaluated during the years of 1381- 1387. Finally, the optimal forecasting model is proposed as a multi input- multi output neural network structure with a novel algorithm. The performance of the proposed structure is evaluated during the years of 1388- 1393 in the forecasting process.