A Machine Learning Approach for Forecasting Energy Use in the Transportation Sector of the USA

Transportation is an important sector for any country and its economy. The sector uses energy from different available sources. Due to the extremely high volatility in energy demand, predicting energy use patterns from various sources is difficult. This study considers three major sources of energy used in the transportation sector of the USA. They are Natural Gas (NG), Petroleum (PT), and Fossil Fuel (FF). We perform a predictive modeling exercise based on monthly data collected from January 1973 to August 2023. We use Extreme Gradient Boosting (XGBoost), an ensemble Machine Learning (ML) approach. The results indicate that the consumption of NG is much less compared to FF. The consumption of NG started to increase during the last four years. Among the three sources, energy used from FF is the most predictable, followed by PT and NG. The findings of this study will be useful for practitioners and investors in the field of energy used in the transportation sector of the USA.