<![CDATA[This study aims to develop and evaluate a stock price prediction model for Bank Central Asia (BBCA) using a multivariate Long Short Term Memory (LSTM) approach. The model utilizes four key variables from historical stock data open, high, low, and close prices and is compared to a univariate model that uses only the closing price as input. The research process includes data preprocessing, LSTM architecture design, model training over 200 epochs, and performance evaluation using MAE, RMSE, and MAPE metrics. The results demonstrate that the multivariate LSTM model provides higher predictive accuracy, achieving a MAPE of 2.41%, outperforming the univariate model which recorded 2.71%. Moreover, the multivariate model shows better stability across validation and test data, and greater adaptability in capturing market dynamics. Prediction result visualizations support these findings, with the multivariate model producing more consistent forecasts that closely follow actual data. These results suggest that integrating OHLC variables enhances prediction accuracy and model reliability. This study contributes to the advancement of stock price prediction systems based on deep learning and serves as a valuable reference for investors and decision-makers in designing more data-driven investment strategies.]]>
Copyrights © 2025
