Volumetric Primacy in Energy-Efficient Residential Design: A Multi-Method Framework for Mediterranean Climates

Energy consumption in residential buildings is largely driven by heating, ventilation, and air conditioning (HVAC) systems, which account for more than 50% of total building energy use. Although conventional energy-efficiency strategies primarily emphasize fenestration improvements, the relative influence of architectural design parameters on thermal loads in Mediterranean climates remains insufficiently understood. This study identifies the building characteristics that most strongly influence energy performance under these climatic conditions. The dataset, originally published by Tsanas and Xifata (2012), comprises 768 simulated residential building configurations representative of Athens, Greece. This study identifies the buildings characteristics that most strongly influence energy performance under these climatic conditions. A cross-validated multi-method framework integrating statistical feature selection and machine learning was applied to analyze the dataset. The analytical framework combined Spearman's rank correlation, chi-squared testing, principal component analysis (PCA), LASSO regularization, and three supervised machine-learning models: decision trees, artificial neural networks, and support vector regression, optimized through grid-search hyperparameter tuning with cross-validation. PCA showed that volumetric design parameters, particularly building height (loading = - 0.91) and orientation, explain more than 50% of the total variance, whereas glazing characteristics contributed less than 10%. Chi-squared analysis confirmed the dominant influence of orientation (χ² > 400) while indicating that glazing-related variables were not statistically significant (p > 0.05). All machine-learning models achieved excellent predictive performance on the test set (R² > 0.94), with support vector regression producing the highest accuracy (test R² = 0.945). The proposed framework demonstrates that orientation influences energy performance through discrete thresholds, whereas building height exerts a continuous effect. These findings shift energy-efficiency priorities toward climate-responsive volumetric optimization, reducing HVAC energy demand by 18-22% through enhanced stack-effect ventilation and supporting sustainable residential design with lower operational carbon emissions in Mediterranean-climate regions.