Performance Comparison of Machine and Deep Learning Methods in Usd/Try Exchange Rate Forecasting

Abstract

Accurate estimation of exchange rates is very important for economic and financial analysis. Türkiye has been facing serious exchange rate fluctuations, especially recently. At this point, accurate prediction of exchange rates is of great importance for both individual and institutional investors. In this study, 149 months of data between January 2012 and May 2024 were used to estimate the USD/TRY exchange rate. Total Opened USD Deposits, M3 money supply, total imports, total exports, unemployment rate, gold price, CPI, PPI and central bank net dollar reserve were used as input variables in the study. In the study, predictions were made using XGBoost, RandomForest, LightGBM, LSTM and SVR methods. Additionally, the generalizability of the results obtained with five-fold cross-validation was tested. According to the results obtained, the best prediction performance for training, testing and cross-validation data sets was produced by the Random Forest model.