Forecasting Iran's Non-Oil Exports Affected by the Exchange Rate Volatility Using Multilayer Perceptron (MLP) Model

The growing importance of independence from oil revenues due to oil price fluctuations and global demand that severely affect on the government revenues and the economy has led to role of non-oil exports be raised beyond a tool for earning exchange revenues. That's why many economic experts and researchers focus on the analysis of current situation of the nonoil exports. The perspective of the non-oil exports provide the possibility of more accurate survey and planning for Iranian economy. In this study, Artificial Neural Network (Multilayer Perceptron MLP) and Autoregressive Integrated Moving Average (ARIMA) has been used to predict the non-oil exports of Iran during the period 1959- 2010. The needed data has been obtained from the Central Bank of Iran. In order to compare the accuracy of the prediction method several measures (including the mean of absolute deviation, root of mean square error and determination coefficient) has been used. Results showed that Multilayer Perceptron has a lower error in forecasting the non-oil exports and also it significantly was more accurate than ARIMA model. ANN method in addition to create a context for development the new methods of the forecasting, can also help the policy makers of export sector, especially non-oil export sector in the future decisions. Finally, it seems that the managed dollar rate policy in the country is an efficient policy, provided that this rate be not always fixed in the long run.