Forecasting Crude Oil Dynamics Using Multi-Frequency Machine Learning Approaches: Advanced Machine Learning Insights for Energy Market Stability

This study examines the forecasting of crude oil prices using a range of advanced machine learning techniques, including Support Vector Regression (SVR), AdaBoost, Gradient Boosting Regression (GBR), k-Nearest Neighbors (KNN), Neural Networks (NN), and Random Forest (RF), applied across daily, weekly, and monthly time scales spanning June 2010 to December 2024. The analysis integrates descriptive statistics, unit root tests, and Seasonal-Trend decomposition using Loess (STL) to identify underlying structural components and assess stationarity properties, thereby informing model selection and architecture. Model performance is evaluated using multiple metrics, including Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), R-squared (R²), Mean Absolute Percentage Error (MAPE), and Explained Variance Score (EVS), with sensitivity analyses conducted to examine robustness across temporal aggregations. Results indicate that SVR consistently achieves superior predictive accuracy, effectively capturing the non-linear interactions, medium- to long-term trends, and structural shifts characteristic of crude oil price dynamics. These findings underscore the value of multi-frequency, deco