Hybrid LSTM Model for Predicting Indonesian Telecommunication Companies Stock Price

Stock investment is rapidly increasing in Indonesia, focusing significantly on telecommunication stocks. As the market grows, it becomes more critical for investors to analyze stock price movements accurately. However, analyzing stock prices by hand has its challenges. It is difficult to handle large amounts of data and identify complex patterns by hand. Investors increasingly rely on computer-based models to assist them in making predictions to address this problem. This study focuses on developing a Long Short-Term Memory model to analyze the stock prices of Indonesia's two largest telecommunication companies. At first, the model's basic setup did not perform as well as anticipated. To enhance its performance, we applied hyperparameter tuning, which significantly improved the model's accuracy. The best-tuned Long Short-Term Memory model achieved a Root Mean Square Error of 1.81366% and MSE of 0.03289%. Additionally, we experimented with a hybrid model that combined Long Short-Term Memory with eXtreme Gradient Boosting. This approach further reduced the Root Mean Square Error to 1.80583% and the Mean Square Error to 0.03261%, demonstrating even better performance. These findings highlight the importance of optimizing models to improve stock price predictions, providing investors with more reliable and accurate information to make well-informed decisions in the stock market.