<![CDATA[The high volatility and nonlinear dynamics of Antam gold prices present significant challenges for accurate time series forecasting, particularly within emerging financial markets. This study aims to develop and evaluate a comparative forecasting framework by examining the performance of the Nonlinear Autoregressive Neural Network (NARNN) and the Holt–Winters exponential smoothing method. A quantitative approach was applied using daily gold price data from January 4, 2010, to January 4, 2025. Data preprocessing included linear interpolation for missing values, Box–Cox transformation for variance stabilization, and time series decomposition to identify structural patterns. The dataset was partitioned into training and testing sets using an 80:20 ratio. Model performance was assessed using the Mean Absolute Percentage Error (MAPE). The results demonstrate that the NARNN model significantly outperforms the Holt–Winters approach, achieving a MAPE of 0.44%, compared to 11.43% and 11.90% for the additive and multiplicative variants, respectively. These findings highlight the limitations of classical linear smoothing methods in capturing abrupt structural changes and confirm the superiority of nonlinear neural network models in modeling complex financial time series. This study provides a robust empirical contribution by establishing a comparative modeling framework that enhances forecasting accuracy in volatile commodity markets.]]>
Copyrights © 2026
