Forecast of Total Factor Productivity in the Economy

The main purpose of study is to predict the total factor productivity in Iranian economy and it has been done by Using artificial neural networks and descriptive-sectional study. In this study, total factor productivity parameters Output in Iran such as Inflation, real exchange rate, foreign debt, education and human capital, foreign direct investment, economic openness are considered as input for Neural networks and the output of the neural network is total factor productivity in Iranian economy. The time period examined has been 1996 to 2016 years. In this study, the feed forward neural network whit back propagation algorithm used to predict. 5 different configurations of inputs are used to design 5 models and 30 different scenarios in each model on the number neurons, designed and implemented to predict. Best models of neurons and networks of neurons fourth model with 19 active functions TANSIG input and output function is LOGSIG and by correlation coefficient (R) Mean Square Error (MSE) and root mean square error (RMSE) and normalized root mean square error (RMSE N), a Best model has been selected and they are equal to 0.9985,0.0111,0.1055 and 2.62, respectively. This indicates that the artificial neural networks were designed in this study have the ability to predict the Iranian economy's total factor productivity.