---

The objective of this research work is to forecast the rice and corn imports using the methods of artificial neural network and ARIMA. The results are then compared. Calculations are based upon the data from the Iranian Customs regarding imports of rice and corn during the period 1360 to 1383. Data for the period 1360-1380 are used for educating the network, and data related to the last three years are used to examine the predictive power. The results show that the method of neural network has a better performance as compared to the process of ARIMA and is able to predict the rice and corn imports with higher accuracy.

---

This study is carried out with the objective of forecasting nominal and real prices of some agricultural products such as onions, potatoes and tomatoes. After the analysis of stationarity of the associated time-series, the randomness of the variables is studied using the non-parametric Wald-Wulfowitz and the parametric Durbin-Watson tests. The results show that all the nominal price series of the said products and the real price series of potatoes were identified as non-random and forecastable time-series. The time-horizon for this study is 1350-1384. Models used for the forecasting in this analysis were AR, MA, ARIMA, Single Exponential Smoothing, Double Exponential Smoothing, Harmonic Analysis, ARCH and the Artificial Neuro-network. Amongst the four forecasting series based on the minimum-error criterion, the AR and MA models forecasted the nominal price series of onions and potatoes much better. The Harmonic model forecasted the nominal price of tomatoes with minimal error. The forecasted values for the real price of potatoes using the ARCH model produced the minimum errors. Forecasting error for the nominal price series of potatoes was less than its real price series. The forecasting values for the year 1385 and 1386 were shown to be mere than the similar values for 1384, with the exception of the price for potatoes.