The development of the new protectionism and investigation of interaction with import by artificial neural networks method: The case of Turkey

The aim of this study; The aim of this study is to examine the usage areas of the new protectionism in free trade and to determine the best architecture to predict imports by using the artificial neural network model and to measure the interaction of the new protectionism on imports. In this context, the development of the new protectionism in the historical process was examined, and the interaction of the new protectionism on imports was compared with other macroeconomic variables. In the analysis part, Turkey's quarterly data for the years 1998-2020, Gross Domestic Product by Expenditure Method, CPI-based Real Effective Exchange Rate, export-import coverage ratio and new protectionism variables were used as determining variables. Different architectures were created using the artificial neural network model and the weights of the variables were determined. The most appropriate artificial neural network architecture was created by comparing the estimated data that emerged after the determination of the weights of the variables and the actual data. The most appropriate artificial neural network architecture was created by comparing the estimated data that emerged after the determination of the weights of the variables and the actual data. It has been determined that the artificial neural network architecture created has a high explanatory power with a ratio of 0.95. The actual data and the estimated data were compared by calculating R2, RMSE and VAF values. In experiments with 5 neuron cells in the model, R2=0.95, RMSE=0.07 and VAF= 95.09. In order to predict the output in the learning process, the coefficient weights of the inputs were calculated. The coefficient of the new protectionism tools in the model was 0.672724, the GDP coefficient was-0.08987, the coefficient of the real exchange rate was-0.70297, and the coefficient of the export-import coverage ratio was 0.373448. The results showed that the explanatory power of new protectionism tools is high among the variables used in import estimation, the predictability of the created artificial neural network model is consistent and its reliability is high.