Neural networks with value of Turkey's GDP forecast

The Gross Domestic Product (GDP), which is one of the most important instruments of national income that measures the level of economic growth and economic development, provides information about the general outlook of the countries. Economies with high GDP figures are considered advanced. Also for Turkey's economy will contribute to the development of the rise in these figures. The aim of this study was to estimate the value of Turkey's GDP. In this context, firstly, Structural Equation Model and Partial Least Squares Method were applied by selecting the variables that could affect the GDP between 1998 and 2017 and then estimating the GDP with the Artificial Neural Network Model. According to the results of the Structural Equation Model, the selected variables explained the GDP with a high value. In the model, fixed capital investment-total domestic savings and consumption variables were supported, while other variables were not supported. In addition, a high value of the goodness of fit index shows that the study has universal validity. Looking at the results of the Artificial Neural Network model which is created by trying the hidden layer number between 2 and 5, the best result is reached when the number of hidden layers is selected as 5. It was concluded that the model established by obtaining the values of R2 = 0,996140651, RMSE = 19444911,6, MAE = 15845918,2 and MAPE = 32,29791086 in 5 hidden layers is acceptable. In addition, as the number of hidden layers increased, the model showed better results. The real values of GDP and estimation values are very close to each other. It is determined that the Artificial Neural Network has a high predictive power. Keywords: Structural Equation Model, Partial Least Squares, Artificial Neural Network, GDP