A Hybrid Framework Combining Data-Driven and Expert-Driven Approaches in Building Energy Modeling based on simulation.

Buildings account for 39% of global energy use and 38% of greenhouse gas emissions, highlighting the need for improved energy analysis. This study proposes a hybrid framework integrating machine learning and multi-criteria decision-making to predict and prioritize factors affecting building energy consumption. A real office building in Mehrabad, Tehran, was modeled in EnergyPlus, generating 4,200 monthly datasets through parametric simulation. After Lasso-based feature selection, XGBoost, Random Forest, and SVM were trained, with XGBoost achieving the highest accuracy (RMSE = 215.62, R² = 0.99). SHAP (data-driven) and AHP (expert-driven) were then applied to rank eight key features. While partial agreement was observed, notable differences highlighted the complementary strengths of both approaches. Integrating SHAP and AHP provides a balanced and explainable prioritization framework, offering a novel contribution to data-informed and expert-supported decision-making in sustainable building design and operation.