Evaluating regional sustainable energy potential through hierarchical clustering and machine learning

Abstract Energy is an essential resource for sustaining daily life and achieving economic growth. The increase in global energy demand, combined with the adverse environmental impacts of fossil fuels, has highlighted the urgency of transitioning to sustainable energy sources. In large and heterogeneous countries like Türkiye, region-specific analyses of sustainable energy potential (SEP) are crucial for formulating effective policies and optimizing resource allocation. This study introduces a novel two-step hierarchical clustering and classification framework to evaluate the SEP of Türkiye’s provinces comprehensively. The framework combines fuzzy and crisp clustering methods to capture the complex relationships among socioeconomic, geographical, and renewable energy potential criteria. First, the Fuzzy C-Means (FCM) algorithm is employed to perform fuzzy clustering using three main criteria and three different distance metrics—Euclidean, Manhattan, and Minkowski—, resulting in 21 clustering scenarios. Membership degrees from the fuzzy clustering phase are then integrated into a new dataset, which undergoes crisp clustering using the K-Means algorithm. This approach provides both granular and definitive cluster structures, enabling a robust analysis of regional energy characteristics. To validate the clustering results, supervised classification methods—including K-Nearest Neighbors (KNN), Support Vector Machines (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGBoost)—are utilized, alongside ensemble models based on RF and XGBoost. The classification results are compared with traditional clustering evaluation indices, such as the Silhouette and Calinski-Harabasz indices, demonstrating the feasibility of using classification models to assess clustering accuracy. This study’s key contributions lie in integrating clustering and classification methods systematically and providing actionable insights into which renewable energy sources are most suitable for each cluster. By tailoring policy recommendations to the unique characteristics of each cluster, this framework not only corroborates existing findings in the literature but also extends them by offering a practical methodology for regional energy planning.