Application of NNARX to Agricultural Economic Variables Forecasting

The aim of this research is studying the application of NNARX as a nonlinear dynamic neural network model in contrast with ARIMA as a linear model to forecast Iran’s agricultural economic variables. As a case study the three horizons (1, 2 and 4 week ahead) of Iran’s rice, poultry and egg retail price are forecasted using the two mentioned models. The results of using the three forecast evaluation criteria (R2, MAD and RMSE) state that, NNARX model outperforms ARIMA model in agricultural economic variables forecasting.