Predicting energy commodity prices using machine learning: A study of crude oil and natural gas price dynamics

This paper aims to predict the commodity price of crude oil and natural gas by applying Bayesian Ridge regression model. The focus is on evaluating the model’s ability to give actionable insights for investors in the commodity markets. The research implements the Bayesian Ridge regression, which is a linear model that combines the regularization and Bayesian to predict future prices of crude oil and natural gas. Historical price data for crude oil and natural gas are used as input features, where the model trained on this data to generate predictions. The model is evaluated using evaluation metrics including Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE) and Coefficient of Determination (R²). A comparison is done with the two commodities aiming on how well the model captures historical price movements and its ability to predict future trends for the investor’s decision making. The model shows high predictive capabilities for both commodities with high R² scores whereas, it shows low error metrics for natural gas achieves reasonable accuracy for crude oil. The findings recommend the model is a valuable tool for identifying trends especially for natural gas, which shows stable patterns that was predicted with high accuracy. This paper concludes, that Bayesian Ridge regression model is a robust tool for predicting mainstream movements in commodity prices.