Forecasting inflation in Turkey: A comparison of time-series and machine learning models

Purpose - This paper aims to analyze inflation in Turkey. For this purpose, the accuracy of some Machine Learning (ML) models in forecasting inflation have been tested a new and complementary approach has been tried to be given to time series models.Methods - This paper forecasts inflation in Turkey by using time-series and machine learning (ML) models. The data spans from the period 2006:M1 to 2020:M12.Findings -According to the root mean squared error and R-square evaluation criteria, the forecasts obtained from the ML algorithms were less accurate than the forecasts obtained from the VAR model. However, it has been observed that the findings obtained from the MLP algorithm, which takes into account nonlinear relationships, give more accurate results compared to the forecasts obtained from linear-based Lasso and Ridge models. From this point of view, it is suggested that nonlinear ML should be evaluated as a complementary method for inflation forecasting.Implication - According to the study's findings, the nonlinear ML algorithms can be thought of as a complementary method to forecast inflation in emerging economies with volatile inflation rates. Central banks and policymakers can benefit from computational power and big data for inflation forecasting.Originality - We evaluate the forecasting performance of ML models against each other and a time series model and investigate possible improvements upon the naive model. So, this is the first study in the field that uses both linear and nonlinear ML methods to compare the time series inflation forecasts for Turkey.