<![CDATA[Stock price prediction is a highly complex problem in the world of finance because it is influenced by various nonlinear and dynamic factors. This study compares the performance of three deep learning models Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Recurrent Neural Network (RNN) in predicting the closing price of PT Bank Rakyat Indonesia (Persero) Tbk. (BBRI) stock, with the ticker symbol BBRI.JK. The dataset used consists of historical BBRI stock price data from January 2, 2014, to December 30, 2024, covering 2,719 active trading days. The data was divided into 80% training data (2,115 sequences) and 20% test data (544 sequences) with a lookback window length of 60 days. Model evaluation was performed using three primary metrics: Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE). The experimental results show that the LSTM model provides the best prediction accuracy with a MAPE of 3.8722%, an RMSE of 194.25 IDR, and an MAE of 162.22 IDR, followed by the RNN model (MAPE 9.8298%, RMSE 493.57 IDR, MAE 417.36 IDR), and the CNN, which had the lowest performance (MAPE 25.3342%, RMSE 1,060.64 IDR, MAE 1,030.64 IDR). The LSTM’s coefficient of determination (R²) of 0.9531 confirms its ability to capture long-term patterns in stock time series data. This study recommends the LSTM model as the preferred choice for implementing a stock price prediction system for the banking sector in Indonesia.]]>
