IMPLEMENTASI LONG SHORT-TERM MEMORY DALAM MEMPREDIKSI PERGERAKAN HARGA SAHAM DI BURSA EFEK INDONESIA

Stock investment offers high return potential but is highly vulnerable to risk due to dynamic market volatility. This highlights the importance of accurate stock price
prediction in supporting investment decision-making. This study aims to implement a Long Short-Term Memory (LSTM) to forecast the stock prices of five banking sector companies listed in the LQ45 index of the Indonesia Stock Exchange: BBCA, BBNI, BBRI, BBTN, and BMRI. Daily historical data from January 2, 2020, to December 30, 2024, were collected through web scraping from Yahoo Finance. The
dataset was preprocessed using a windowing technique (n_past = 30, n_future = 1) and normalized with RobustScaler. The model was configured with Dropout, Batch Normalization, L2 Regularization, and ReLU activation, and trained using the Adam optimizer with a ReduceLROnPlateau callback. Hyperparameter tuning was conducted using Keras Tuner with the Random Search method. Evaluation on
the test set yielded an RMSE of 156.23 and MAE of 111.35. Residual analysis revealed evenly distributed errors with minimal bias, and an average relative error below 5%. These results indicate that the proposed model provides stable and reliable stock price predictions based on historical time series data.