<![CDATA[Gold is a strategic asset widely used as a hedging instrument in Indonesia, with its price influenced by domestic macroeconomic factors including inflation, Bank Indonesia's benchmark interest rate, and the USD/IDR exchange rate. Classical statistical models such as ARIMA have limitations in capturing non-linear patterns and long-term dependencies in financial time series data. This study compares the performance of two deep learning algorithms, Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU), in predicting the daily price of ANTAM LM gold based on the aforementioned macroeconomic variables. Data covering the period from January 2015 to September 2025 were obtained from PT Antam, Bank Indonesia, and Investing.com. The research process included data collection, preprocessing, exploratory data analysis, dataset splitting under three schemes (70:30, 80:20, and 90:10), model training with hyperparameter tuning, and evaluation using RMSE and MAPE metrics. Results demonstrated that GRU achieved the best performance under the 90:10 split scheme with an RMSE of IDR 35,417.91 and a MAPE of 1.47%, outperforming the best LSTM model which recorded an RMSE of IDR 96,808.45 and a MAPE of 4.58%. The best-performing GRU model was deployed into a Flask and React-based web application serving as a tool for investment decision-making.]]>
Copyrights © 2026
