Prediction of Indonesian Coal Prices Using Support Vector Regression Method with Grid Search Optimization

บทคัดย่อ

Coal prices are highly volatile due to fluctuations in global supply and demand, government policies, and economic and political factors. Therefore, accurate price prediction is essential and can be achieved using Support Vector Regression (SVR). This study aimed to identify the most accurate SVR model for forecasting Indonesian coal prices and analyze its prediction performance. The Radial Basis Function (RBF) kernel was used with hyperparameter ranges of 10⁻³ ≤ 𝐶 ≤ 10², 10⁻³ ≤ 𝛾 ≤ 10², and 𝜀 = 0.01, 0.02, 0.03, tested at data splits of 70:30, 80:20, and 90:10. Optimal hyperparameters were determined using the grid search algorithm, and model performance was evaluated using the Mean Absolute Percentage Error (MAPE). Results showed that the best prediction accuracy was achieved with a MAPE of 5.459%, indicating excellent model performance. The optimal configuration was the SVR with RBF kernel at a 90:10 ratio, where 𝐶 = 12.5, 𝛾 = 0.1, and 𝜀 = 0.025. The resulting model effectively approximated the actual coal price data, confirming its reliability for forecasting Indonesian coal prices.