Perbandingan Metode Extreme Gradient Boosting (XGBOOST) Dengan Long Short-Term Memory (LSTM) Untuk Prediksi Saham Pt. Bank Mandiri Tbk. (BMRI)

Abstract

In the continuously evolving world of investments, achieving optimal investment results is the primary goal of every investor. The high profit potential in stock investments makes it an attractive choice. However, it is difficult to predict the direction of stock price movements, but many methods and ways are used to predict in terms of buying and selling, one of which is the result of the very rapid development of computing for machine learning, namely artificial intelligence techniques. In this study, the artificial intelligence methods used are Extreme Gradient Boosting (XGBoost) and Long Short-Term Memory (LSTM). This study uses a stock dataset of PT. Bank Mandiri Tbk. (BMRI) for 20 years, especially in the open price column. After testing, the results from the XGBoost method are obtained, namely the Coefficient of Determination (R2) value of 89.09%, indicating that the results are good and the Mean Absolute Percentage Error (MAPE) is 3.21%, indicating that the error percentage is low. While in the LSTM method, the R2 value is 98.44%, meaning that the prediction results have been predicted very well and the MAPE is 1.77%, indicating that the error percentage is very low.

Keywords: XGBoost, Open Price, Prediction, LSTM, Stocks