Turkey's macroeconomic performance forecast by (with) the method of artificial neural networks

Inflation, unemployment, GDP, Budget Deficit, Current account deficit, which are accepted as basic variables in terms of macro economics, are of great importance for the economies of countries. Therefore, these variables are also considered as basic criteria in measuring the economic performance of a country. Accordingly, the fluctuations in the economy primarily affect this variable, indicating the importance of variables for the stability of the country's economy. Inflation, unemployment rate, GDP, budget deficit, the trend of current account rations are primary parameters, which are observed while evaluating Turkey's economy. Artificial neural networks method works like a human brain. The brain makes new decisions and makes new forecasting in the light of the information it has learned in a sample event. In the method of artificial neural networks, the network is first trained just like the brain by givinghistorical data and the learning of the network is provided. The learning network can now make its own forecasting. In this study, a statistical estimation techniques, which is artificial neural network, is used to estimate macroeconomic performance of Turkey by using basic economic indicators and data between 1990-2018. Firstly, the data of factors that affect the Turkey's economy is firstly gathered and analyzed. Then, ANN model was created based on these factors and the model was trained and tested. Afterwards, annual macroeconomic performance is estimated with the model and performance tests of these predictions was calculated. As a result of the study, the performance of the ANN predicction model was measured as the correlation coefficient 0.99 and the regression coefficient 0.98. In other words, when the estimates produced by the model were compared with the actual values, it was concluded that the values of MSE, RMSE and MAPE were low and therefore the obtained model produced slightly erroneous results and the prediction consistency was high. In addiction, this study provides an opportunity to increase the recognition and usability of ANN in studies to be carried out in our country by the researchers by introducing ANN, which has been used quite recently in the literature.