<![CDATA[Gold price volatility poses a major challenge for investors in making accurate and timely investment decisions, as price movements are influenced by complex and nonlinear financial dynamics. Conventional forecasting models often fail to capture these patterns optimally due to suboptimal parameter selection. This study aims to develop a hybrid forecasting model by optimizing the Long Short-Term Memory (LSTM) network using a Genetic Algorithm (GA) to improve gold price prediction accuracy. Historical gold price data spanning from 2004 to 2025, including Open, High, Low, and Close attributes, were utilized in this research. The methodological framework consists of data preprocessing, normalization, time-series sequence construction, dataset partitioning using an 80:20 training–testing ratio, and hyperparameter optimization through GA. Model performance was evaluated using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and prediction accuracy. Experimental results demonstrate that GA-based optimization significantly enhances LSTM performance, with RMSE reduced by more than 80%, accompanied by substantial decreases in MAE and MAPE, and an increase in prediction accuracy to 99.77%. These findings confirm that systematic hyperparameter optimization plays a crucial role in improving deep learning model generalization for financial time series forecasting. The proposed LSTM–GA model contributes to the development of more robust and reliable predictive frameworks for financial applications, particularly in supporting investment decision-making under volatile market conditions.]]>
