Data on forecasting energy prices using machine learning.

Area of Science:

• Energy Economics
• Data Science
• Commodity Markets

Background:

• Accurate energy commodity price forecasting is crucial for market stability and economic planning.
• Traditional econometric models often struggle with the complex dynamics of energy markets.
• Machine learning offers advanced capabilities for analyzing large, time-series datasets.

Purpose of the Study:

• To present a comprehensive dataset for the long-term forecasting of energy commodity prices.
• To facilitate the application and comparison of various predictive modeling techniques.
• To serve as a benchmark for future energy price prediction research.

Main Methods:

• Utilized monthly price data for six major energy commodities spanning nearly four decades.
• Applied a hybrid approach combining econometric models, artificial neural networks, and random forests.
• Employed a standard 80-20% data split for training and testing, with performance evaluated using RMSE, MAPE, and M-DM tests.

Main Results:

• The dataset enables the evaluation of multiple forecasting methodologies on historical energy commodity prices.
• Comparative analysis of traditional and machine learning models provides insights into their predictive accuracy.
• Established a robust dataset for benchmarking new forecasting algorithms in the energy sector.

Conclusions:

• The presented data is valuable for researchers and practitioners in energy economics and data science.
• Machine learning models demonstrate potential for improving long-term energy price forecasting accuracy.
• The dataset supports further investigation into advanced forecasting techniques and market dynamics.