Rekayasa Fitur dan Gradient Boosting untuk Prediksi Harga Saham Pada Pasar Saham Indonesia

Abstract

This study aims to analyze the comparative performance of three machine learning models Neural Network, Random Forest, and XGBoost in predicting the stock price of Bank Rakyat Indonesia (BBRI.JK) based on feature engineering integration. The background of this study is based on the need to develop accurate and efficient predictive models to deal with stock market volatility. The Data used covers the period 2010-2025 with the application of technical indicators such as Moving Average (MA), Relative Strength Index (RSI), volatility, and price momentum as the main features. The research method uses a machine learning approach based on supervised learning with a five-fold cross validation process. Model evaluation was conducted using quantitative metrics including Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), coefficient of determination (R²), and Mean Absolute Percentage Error (MAPE). The results showed that XGBoost produced the Best Performance With R² = 0.9451, MAE = 87.3129,and MSE = 10327.1187, followed by Random Forest (R² = 0.9233) and Neural Network (R² = 0.9120). The XGBoost Model proved to be the most stable and efficient in handling nonlinear data as well as extreme price fluctuations. The discussion confirms that the integration of engineering features improves the generalization capability of the model and lowers the prediction error rate significantly. Future research is recommended to include macroeconomic variables, sentiment data, and reinforcement learning approaches to broaden the scope and improve the model's adaptability to global financial market dynamics.