<![CDATA[Stock investment decisions in emerging markets are often hindered by high volatility and non-linear price patterns. Conventional statistical methods frequently fail to capture the long-term dependencies inherent in financial time series data. This study proposes a data-driven forecasting framework using Long Short-Term Memory (LSTM) optimized with Hyperparameter Grid Search to predict the closing prices of Indonesia's top capitalization banking stocks: BBCA, BBNI, and BMRI. The methodology involves comprehensive data preprocessing, including MinMax normalization and an 80:20 train-test split, followed by a rigorous hyperparameter tuning process to determine the optimal epochs and batch sizes. The model's performance was evaluated using Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE). The results demonstrate that the proposed LSTM framework achieves high forecasting accuracy. Specifically, the model yielded a MAPE of 0.99% for BBCA, 1.35% for BBNI, and 1.38% for BMRI, indicating a very low error rate. These findings confirm that the optimized LSTM model can effectively capture dynamic market trends, providing investors and portfolio managers with a reliable decision support tool for strategic asset allocation.]]>
Copyrights © 2026
