Analysis of savings gap with neural network model in Turkey

The main objective of this thesis is to determine the best architecture to estimate the saving gap by using the artificial neural network method and to examine the changes in the historical process of saving, investment and saving gap in our country. In this context, the saving gap was analyzed structurally, the relationship between the other economic variables was examined and the predictability of the saving gap with artificial neural networks was performed. Although artificial neural networks do not solve many problems, no matter how much they learn, they still make mistakes. Artificial neural networks behave according to their own decisions. To estimate Turkey's saving gap data in this study, the rate of inflation in the past year, nominal interest rate, gross external debt per capita GDP and public expenditure data are used as training data of Neural Networks. Artificial neural networks were created by using Matlab Neural Network toolbox. Matlab environment; Network type: Feed-forward beckpropogation, Training function: Trainlm (Levenberg-Marquardt), Adaptive learning function: Learngdm, Performance function: mse (Mean Squared Error) neural networks training was carried out. Neural network trained with Turkey for the past year savings estimate of the saving gap data was performed. The determination of the data which is not used in education with high accuracy shows that the training of artificial neural network is successful.