BLDAR: A Blending Ensemble Learning Approach for Primary Energy Consumption Analysis

The ability of a country to control its primary energy consumption serves as vital for attaining development, environmental sustainability, and economic stability. From national security and economic success to global climate changes and geopolitical alliances, a country’s primary energy consumption has a significant impact on a wide range of problems that affect both its internal well-being and the worldwide landscape. Analysis of primary energy consumption is indispensable for the development of both short-term and long-term national strategies, and precise prediction of future consumption patterns is of great importance for successful decision-making at the root level. In this study, we use several Machine Learning (ML) algorithms for forecasting the primary energy consumption based on related Sustainable Development Goals (SDGs) variables. Specifically, we developed a forecasting model using a blended ensemble learning model, namely blending LDAR that blends Light Gradient Boosting (LGB), Decision Tree (DT), AdaBoost (ADB), and Random Forest (RF). We utilize global data on sustainable energy at various time series frequencies and different training and testing ratios to evaluate our proposed LDAR and other ML models. Our proposed model achieved significantly high performance in each ratio of training and testing. Proposed LDAR performs better than other ML models and achieves 0.0177 MSE, 0.0016 MAE, 0.0403 RMSE, 19.7075 SMAPE, and 90% R 2 $$R^2$$ score. Proposed models help the policymakers and stockholders to achieve SDGs in terms of energy consumption. Future research focuses to integrate related SDG features and bring more dynamic models in analysis with model explainability.