Comparing Forecasting Ability of Artificial Neural Networks and ARIMA Methods in Forecasting of Iran's Leather and Skin Exports

Forecasting economic variables as a helpful planning tool has got considerable attention in economic literature. There are various methods that could be used for forecasting. The objective of this study is to identify an efficient method for predicting Iran’s leather and skin exports. For this purpose، auto-regressive integrated moving average process (ARIMA)، artificial neural networks (ANN) and hybrid methodology of ANN-ARIMA were compared using time series data from 1971 to 2010. The results showed that hybrid methodology of ANN-ARIMA that decompose the exports time series into its linear and nonlinear forms has better forecasting performance and higher accuracy compare to other methods. In addition، the hybrid methodology as the most accurate method in this research، forecasted the amount of leather and skin export better than the other two methods. However، similar to other two methods it predicts that leather and skin exports would have a downward trend in coming years.