Accurate forecasting of Islamic stock prices is essential for supporting investment decisions in increasingly dynamic capital markets. However, traditional statistical approaches often face limitations in capturing the non-linear and temporal characteristics of stock market data. This study aims to improve the quality of Islamic stock price prediction within the Indonesian Islamic capital market using a Long Short-Term Memory (LSTM)-based deep learning model. Bank Syariah Indonesia (BRIS.JK), one of the largest Shariah-compliant banking stocks in Indonesia, was selected as the research object. The study employed a quantitative experimental approach using 1,504 daily stock observations collected from Yahoo Finance over the period 2018-2024. Data preprocessing included cleaning, Min-Max Scaling, and an 80:20 chronological train-test split. The forecasting model was developed using an optimized stacked LSTM architecture and evaluated using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE). To assess model robustness, the training and evaluation process was repeated ten times under identical experimental settings. The results show that the proposed model achieved an average MAE of 1.04%, RMSE of 1.54%, and MAPE of 1.97%, indicating highly accurate forecasting performance and stable results across repeated experiments. The findings demonstrate that LSTM effectively captures complex temporal patterns in Islamic stock price movements and provides reliable forecasting outcomes. This study contributes to the literature by providing empirical evidence from BRIS.JK, incorporating systematic hyperparameter optimization, and evaluating model consistency through repeated experimental procedures. The proposed model offers a practical decision-support tool for investors and market participants in the Indonesian Islamic capital market.