<![CDATA[Accurate gold price prediction is crucial in supporting financial and investment decision-making. This study aims to develop and optimize a daily gold price prediction model using the Extreme Gradient Boosting (XGBoost) algorithm based on historical price data and technical indicators. The model was constructed to predict two types of prices, namely "Close" and "Buyback" prices in IDR/gram. Optimization was carried out using Bayesian Optimization to obtain the best hyperparameter combinations. The model was evaluated using a Walk Forward Validation (WFV) approach with a 14-day sliding window and two main evaluation metrics: Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). The results show that the model provides excellent predictive performance, with an average RMSE of 15,431.92 and MAPE of 1.03% for Close price, and RMSE of 15,382.64 and MAPE of 1.15% for Buyback price. The prediction visualizations indicate that the model consistently follows the actual price trend. Feature importance analysis reveals that technical indicators such as RSI, EMA, and MACD significantly contribute to the model. The success of this study demonstrates that an optimized XGBoost model can serve as a reliable approach for gold price forecasting and opens opportunities for developing more advanced predictive models in future research.]]>
Copyrights © 2025
