Döviz Kurunu Belirleyen Değişkenlerin Tahmini: Doğrusal Zaman Serisi Ve Doğrusal Olmayan Yapay Sinir Ağı (YSA) İle Bir Model Önerisi

Authors : Ersin Sünbül, Yasemin Keskin Benli

Pages : 192-209

View : 70 | Download : 175

Publication Date : 2025-07-16

Article Type : Research Paper

Abstract :The aim of this study is to identify the most significant variables influencing the exchange rate and to propose a high-performing predictive model using both linear and nonlinear time series methods. A dataset comprising 16 variables obtained from the Central Bank of the Republic of Turkey was analyzed. Three different models were applied: a traditional Artificial Neural Network (ANN) model (97.2% accuracy, RMSE 2.7960), a combined VAR-ANN model (99.6% accuracy, RMSE 0.3184), and a Multi-Stage Data Cleaning–Causality–ANN model (99.7% accuracy, RMSE 0.3062). The third model demonstrated the highest predictive accuracy. The key determinants of the exchange rate were identified as the consumer price index, real interest rate, unemployment rate, current account balance, gross domestic product, and exports. The findings offer strategic insights for policymakers and researchers, contributing meaningfully to the existing literature.
Keywords : Döviz Kuru Teorileri, Toda-Yamamoto Nedensellik Analizi, Yapay Sinir Ağı.

ORIGINAL ARTICLE URL