"Simulation-Driven Xgboost Model for Multi-Output Building Facade Performance Prediction" presents a research study that couples artificial intelligence (AI) techniques with building performance simulations for facade design analysis and optimization. Building performance simulations have become the most widely adopted approach for evaluating facade performance during the design process; however, they are computationally expensive, especially when exploring large design options. To address this limitation, this paper presents a framework for rapid facade performance prediction. First, a Python-based parameter tree was used to generate 11,760 facade design scenarios by combining six design variables. Next, all scenarios were simulated using EnergyPlus and Eppy to derive nine annual facade performance outputs, spanning energy use, comfort, and cost metrics. After data screening, 10,006 scenarios were retained and used to develop a multi-output Extreme Gradient Boosting (XGBoost) model. The results show that the XGBoost model demonstrated excellent predictive performance, achieving test R² values between 0.978 and 0.999, with %RMSE below 6.595% across all nine outputs. Finally, permutation feature importance was analyzed to determine the influence of facade design variables on each performance output. The findings showed that solar control and Window-to-Wall Ration (WWR) were consistently among the most influential predictors across most of the performance outcomes.

Citation:

Aghimien, E., and Aksamija, A., (2026). “Simulation-Driven XGBoost Model for Multi-Output Building Facade Performance Prediction”, Proceedings of the 2026 Annual Modeling and Simulation Conference (ANNSIM), Orlando, Florida, May 4-7.