Application of Bidirectional Gated Recurrent Unit and Stacked Gated Recurrent Unit Algorithms for Stock Price Prediction

Stocks are financial instruments that provide ownership rights and profits to their holders in public companies, which attract investors because of the potential for high returns despite the risks. Because stock price volatility is difficult to predict, accurate analysis techniques are needed to help investors make the right decisions, especially in the banking sector such as Bank Rakyat Indonesia. This study predicts the stock price of Bank Rakyat Indonesia (BBRI) using deep learning techniques, namely Bidirectional Gated Recurrent Unit (Bi-GRU) and Stacked Gated Recurrent Unit (Stacked GRU), which are chosen because of their ability to capture complex time series data patterns. BBRI stock data from 2018 to 2024 and from 2023 to 2024 were collected, processed, and divided into training and testing data with a ratio of 80:20. The Bi-GRU and Stacked GRU algorithms were optimized with Nadam, Adamax, AdamW, and SGDM, and evaluated using MSE, RMSE, and MAPE. As a result, 7-year data outperform 2-year data with the best algorithm Bi-GRU with Nadam optimization, batch size 8, and learning rate 0.001 giving the best performance with MSE 7283.1848, RMSE 85.3416, MAE 65.0879, and MAPE 1.2421%. The best prediction shows a decrease in BBRI stock prices in November 2024 which indicates a decrease towards the end of the year, thus providing valuable information for companies and investors in anticipating market behavior and making the right investment decisions.