Application of Integrated Neural Network and Input-Output Models in Forecasting Total Production and Final Demand



Forecasting of macroeconomic variables has specific importance in economic topics. Indeed, different models are invented to forecast variables to help economic policy makers in adopting appropriate monetary and fiscal policies. In this paper, the performance of integrated model of Input-Output (IO) and neural network is investigated in forecasting final demand and total production and the results are compared with IO model. At the first step, final demand is estimated by using mean of final demand rates over the period 1365-1375, and then total production is forecasted by using IO model. In the next step, two generalized feed forward neural networks are proposed to forecast final demand and total production of the year 1380. Finally, two models are compared and the hypothesis is evaluated by using MSE, RMSE, MAD, MAPE criteria. The results indicate that the integrated model of IO and neural network outperform IO model in forecasting total production.
JEL Classification: C53, D57, C54