<![CDATA[This study aims to forecast the stock price of Unilever Indonesia using the Long Short-Term Memory (LSTM) model. The dataset consists of weekly stock price data from May 2015 to May 2025, representing a financial time series with nonlinear and dynamic patterns. The LSTM model is employed due to its capability to capture long-term dependencies in sequential data. To evaluate model performance, the dataset is partitioned into three training–testing scenarios, namely 90:10, 80:20, and 70:30. Model accuracy is assessed using the Mean Absolute Percentage Error (MAPE). The results indicate that the best predictive performance is achieved using the 90:10 data split, yielding the lowest MAPE value of 6.973%, which falls into the highly accurate forecasting category. In comparison, the 70:30, and 80:20 scenarios produce higher MAPE values of 13.732% and 17.263% respectively. These findings demonstrate that increasing the proportion of training data significantly improves the performance of the LSTM model in forecasting stock prices. This study highlights the effectiveness of LSTM in modeling financial time series and provides practical insights for data-driven decision-making in stock market analysis.]]>
Copyrights © 2026
