Perbandingan Random Forest dan XGBoost dengan indikator teknikal untuk prediksi harga saham Bank Mandiri

Abstract

The volatility of PT Bank Mandiri (Persero) Tbk stock prices during 2020–2025, influenced by the COVID-19 pandemic and monetary policy dynamics, presents challenges for investors in making optimal investment decisions. This study aims to compare the performance of Random Forest and XGBoost in predicting Bank Mandiri stock prices using technical indicators. The study used 1,152 daily observations after preprocessing and incorporated 10 technical indicators: Simple Moving Average, Exponential Moving Average, Relative Strength Index, Moving Average Convergence/Divergence, Bollinger Bands, Stochastic Oscillator, Average True Range, On-Balance Volume, Rate of Change, and Momentum. The dataset was divided into 80% training (921 observations) and 20% testing (231 observations) using a time-series split. Model performance was evaluated using MAPE, RMSE, MAE, and R². The results show that XGBoost outperformed Random Forest, achieving an RMSE of 161.97 and a MAPE of 2.42%, compared with 182.61 and 2.65%, respectively. XGBoost also demonstrated 38.79% faster computational performance. Feature importance analysis identified the daily low price (61.75%), the closing price (19.55%), and the 5-day SMA (3.68%) as the most influential features. These findings indicate that XGBoost provides more accurate and efficient stock price prediction for volatile Indonesian banking stock data.