ANALISIS MODEL ARIMA DAN XGBOOST DALAM FORECASTING HARGA SAHAM BANK TERBESAR DI INDONESIA

Abstract

The blue-chip banking sector in Indonesia, including BBCA, BBRI, and BBNI, exhibits high volatility and non-linear movements, presenting significant challenges for investors in mitigating financial risk. This study aims to develop and comparatively evaluate the performance of the traditional statistical model, Autoregressive Integrated Moving Average (ARIMA), against the modern machine learning algorithm, Extreme Gradient Boosting (XGBoost), in predicting daily closing stock prices for a one-day horizon (H+1). The research methodology employs a Research and Development (R&D) approach, utilizing historical data for model training and validation. The results demonstrate that the XGBoost model consistently outperforms ARIMA across all tested samples. Specifically, XGBoost achieved a Mean Absolute Percentage Error (MAPE) of 0.55% for BBCA, 1.10% for BBRI, and 0.88% for BBNI, which is significantly lower than ARIMA’s MAPE range of 1.48% to 1.60%. The study concludes that ensemble learning models are more effective at capturing non-linear volatility patterns in the Indonesian capital market and serve as a more reliable decision support tool for investors.