<![CDATA[In recent years, Indonesia's capital market has grown significantly, with the number of investors rising from 3.8 million in 2020 to 12.3 million in January 2024. This study explores the application of the Bi-LSTM model to predict BMRI stock prices by systematically optimizing 75 models to obtain optimal hyperparameters. Unlike prior trial-and-error approaches, this research employs structured hyperparameter exploration using data splits of 70:30, 80:20, and 90:10 to evaluate model accuracy and stability. Results show excellent performance with a MAPE of 2.182% on BMRI’s historical closing prices from January 1, 2021, to July 31, 2024, using a 2-layer Bi-LSTM architecture, batch size 16, and 150 epochs. The findings confirm that an appropriate model can produce highly accurate predictions. This study provides insight into Bi-LSTM modeling in the banking sector, offering valuable references and strategic considerations for investors and stakeholders based on predictive results.]]>
Copyrights © 2025
