Forecasting inflation rates in Turkey with Linear Regression, SARIMA, and LSTM

No Thumbnail Available
Authors
Ayan, Can
Issue Date
2023-08
Degree
MSc in Financial Analytics
Publisher
Dublin Business School
Rights
Abstract
Turkey, categorized under emerging economies, has seen fluctuations in its inflation rates, which have recently been notably high. This prompts the need for robust estimation models that can accurately predict inflationary trends. This study aims to contribute to the existing literature by testing and comparing three distinct inflation forecasting models; multilinear regression, SARIMA, and Long Short-Term Memory (LSTM) in the context of Turkey between 2004 and 2023. By contrasting these models with the CBRT's median market participant survey, and using Root Mean Square Error (RMSE) for model evaluation, the study seeks to identify the most accurate model for predicting inflation in Turkey. The outputs of these study shows that usage of SARIMA and LSTM models together outperforms than the individual models and the benchmark survey. Individual level, SARIMA were performed better to capture extreme fluctuations in time series than others.