Comparative Analysis of LSTM and GRU Models for Indonesian Stock Price Prediction with Integrated Technical and Fundamental Indicators

Stock price prediction is a critical task in financial markets, as stock prices are influenced by various factors. This research aims to predict the stock prices of four prominent companies in Indonesia—Bank BCA (BBCA), Astra International (ASII), PT Indofood Sukses Makmur Tbk (INDF), and Telkom Indonesia (TLKM)—using a Long Short-Term Memory (LSTM) model. The dataset used includes historical stock prices, enriched with technical indicators. Subsequently, experiments were conducted by combining these technical indicators with fundamental indicators to improve accuracy. Other models like Gated Recurrent Unit (GRU), Bidirectional LSTM (BiLSTM), and Bidirectional GRU (BiGRU) are also evaluated using Root Mean Squared Error (RMSE). Experiments demonstrate that the addition of fundamental indicators significantly enhances prediction accuracy. Results show that no single model consistently outperforms others. GRU is best for BBCA and ASII stocks with RMSE values of 91.4 and 80.1, respectively. LSTM is top for INDF with an RMSE of 75.68. BiLSTM excels for TLKM with an RMSE of 46.1. However, statistical tests reveal that LSTM consistently outperforms other models across the evaluated stocks, indicating its reliability for diverse stock forecasting scenarios. This research aims to provide valuable insights for investors by integrating technical and fundamental analysis.