چکیده:
One of the important issues in economics is predicting economic growth, given that accurate prediction of economic growth has significant effects on government economic policymaking and planning, and can, in addition to creating a basis for developing new prediction methods, assist policymakers in future decision-making. Therefore, the aim of this article is to predict Iran's economic growth using three models: neural network, autoregressive moving average, and generalized autoregressive conditional heteroskedasticity. The data of this research includes annual Gross Domestic Product from 1959 to 2011. The results of the three aforementioned methods were compared using the root mean square error and the mean absolute percentage error. The results indicate that the prediction accuracy of the neural network with the forward method is better; through this method, it was determined that real Gross Domestic Product will be increasing in the future, but the economic growth rate will not have a suitable trend in the coming years. Therefore, planning in this regard can be of special importance.
خلاصه ماشینی:
Using neural network approaches and autoregressive models in forecasting Iran's economic growth Nazar Dehmordeh 1, Hajar Esna Ashari 2* Ali Sardar Shahraki 3 Date received: 2013/05/01 Date accepted: 2013/06/01 Abstract One of the important issues in economics is forecasting economic growth, considering that accurate forecasting of economic growth has significant effects on government economic policymaking and planning, and can, in addition to creating a basis for developing new forecasting methods, assist policymakers in decision-making.
Therefore, the aim of this article is to forecast Iran's economic growth using three models: neural network, Autoregressive Integrated Moving Average (ARIMA), and Generalized Autoregressive Conditional Heteroskedasticity (GARCH).
The results indicate that the forecasting accuracy of the neural network method is better than the other method; it was determined through this method that the real Gross Domestic Product will increase in the future, but the economic growth rate will not have a proper trend in the coming years.
Table (6): Forecasting results using the neural network method (Refer to the page image) Source: Research findings 6- Conclusion Forecasting of real Gross Domestic Product was performed using three methods: neural network, Autoregressive Integrated Moving Average (ARIMA), and Generalized Autoregressive Conditional Heteroskedasticity (GARCH).
(Refer to the page image) Chart (1): With selected architecture 11-3-3-1-1 Chart (2) shows the forecasting of Gross Domestic Product using three models: neural network, Autoregressive Integrated Moving Average (ARIMA), and Generalized Autoregressive Conditional Heteroskedasticity (GARCH), which, using the evaluation criteria in Table (5), the best model for forecasting this variable is selected.
(
, 
, 
)
