Predicting energy prices and renewable energy adoption through an optimized tree-based learning framework with...

Area of Science:

• Energy Policy and Economics
• Environmental Science
• Data Science and Machine Learning

Background:

• Global energy consumption analysis is crucial for economic stability and environmental sustainability.
• Predicting energy prices and renewable energy share is complex due to numerous influencing factors.
• Existing models often lack interpretability and robustness in forecasting.

Purpose of the Study:

• To develop and optimize machine learning models for predicting the Energy Price Index and Renewable Energy Share.
• To identify key variables influencing predictive accuracy using sensitivity analysis.
• To provide an interpretable framework for global energy forecasting and policy support.

Main Methods:

• Advanced Machine Learning (ML) regression techniques were employed.
• Metaheuristic algorithms were utilized for optimizing ML models.
• SHAP (SHapley Additive exPlanations) and CAM (Cosine Amplitude Method) were used for sensitivity analysis and model interpretation.

Main Results:

• Fossil fuel dependency and carbon emissions were identified as the most significant predictors for Renewable Energy Share.
• Model optimization using metaheuristic algorithms enhanced predictive accuracy and robustness.
• Sensitivity analyses provided clear quantification of input feature influence on model performance.

Conclusions:

• The study presents a technically rigorous and interpretable framework for global energy forecasting.
• Findings highlight the critical role of consumption intensity and environmental indicators in energy markets.
• The developed methodology supports informed energy policy, sustainability initiatives, and optimized energy system performance.