INVESTIGATION OF THE REPUBLIC OF TURKEY CENTRAL BANK’S GOLD RESERVE WITH HOLT-WINTERS EXPONENTIAL SMOOTHING AND ARTIFICIAL NEURAL NETWORKS / Türkiye Cumhuriyet Merkez Bankas？ Alt？n Rezervinin Holt-Winters üstel Düzleme Y？ntemi Ve Yapay Sinir A？lar？ ？le ？ncelenmesi

The Central Banks carry out various studies to make financial arrangements for their countries. In addition, central banks hold reserves for unexpected needs. However, using wrong forecasting methods in a volatile financial market may cause unexpected consequences. From this point, central banks use prediction methods considering the financial structure of their countries. The economic fluctuations make it difficult to model data with classical statistical methods. In recent years, new modeling techniques are generally superior to classical techniques in modeling nonlinear data. We investigated Central Bank of Turkey’s monthly gold reserves with Artificial Neural Networks (ANN) and Holt-Winters Exponential Smoothing methods. Monthly weighted average reserve amounts ($/million) for the period of December 1987 to May 2017 were used in the study. Additive Holt-Winters Exponential Smoothing model's performance compared with the ANN model. ANN method yielded more successful results in terms of R2, MAPE and RMSE values