터키 경제 불확실성에 관한 예측 모델 비교

A Comparison of Prediction Models for Economic Uncertainty in Turkey: Linear Regression Tree, Decision Tree and Neural Network Models

The main purpose of this article is to identify the optimal predictive model for predicting the economic uncertainty of emerging economies in Turkey. To do this, after obtaining a comprehensive set of related data, the final prediction model was selected by comparing predictive power between three models, such as a linear regression tree model, a decision tree model (bagging tree, random forest), and a neural network model, through a cross-validation technique. Key variables in a model include exchange rates, interest rates, stock markets, bond markets, economic atmosphere, counterparty risk, emerging country risk, and strategic import dependence. A series of economic indicators data for quantitative analysis used data from specialized database companies such as Thomson Reuters Datastream, Worldscope, Bankscope, and Osiris. The main finding is that the predictive power of the random forest model showed a relatively low MSE value compared to other predictive models, indicating its suitability as an optimal prediction model.