Comparison of ARIMA, Random Forest, and Hybrid ARIMA-Random Forest Models in Forecasting Indonesian Crude Oil Prices

Abstract

The price of Indonesian crude oil is highly volatile due to global demand fluctuations, energy policies, and geopolitical tensions, making accurate forecasting challenging. This study compares three forecasting models: ARIMA, Random Forest, and Hybrid ARIMA--Random Forest, to identify the most accurate approach. Model performance was evaluated using Time-Series Cross-Validation (TSCV) and metrics including Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE). The results indicate that the Random Forest model, tuned with \texttt{mtry = 1} and \texttt{ntree = 200}, outperformed both ARIMA and Hybrid ARIMA--Random Forest, achieving the lowest MAPE, MAE, and RMSE values. This suggests that Indonesian crude oil prices during the study period are predominantly non-linear, and Random Forest alone effectively captures these dynamics. Forecasts based on this model indicate a short-term increase in prices from 61.10 USD/Barrel in December 2025 to 64.29 USD/Barrel in March 2026, followed by a slight decline and modest recovery by June 2026. Overall, Random Forest provides a reliable and accurate tool for forecasting Indonesian crude oil prices, offering valuable insights for policymakers, investors, and market participants navigating volatile oil markets.