Currently, building energy models, developed using energy simulation tools, represent the most widely adopted approach for evaluating energy-efficient facade design options and determining optimal solutions. However, these models are time-consuming to develop and computationally intensive to run. To address these limitations, hybrid frameworks that integrate building energy simulation and machine learning (ML) techniques have emerged to streamline the simulation process and aid rapid predictions. Nevertheless, most existing frameworks remain code-based and lack interactive interfaces, limiting their accessibility to non-programmers and design practitioners.

This study introduces an interactive facade performance prediction tool developed by integrating energy simulations and ML. The tool was developed using a four-module workflow comprising Python-based parametric scenario generation, Eppy-driven EnergyPlus simulation, XGBoost ML modeling, and Streamlit web interface development. The framework was implemented using a dataset of 11,760 facade design scenarios generated for a 40 ft by 40 ft generic office building located in Salt Lake City, Utah. Annual outputs from these parametric simulations, consisting of energy use, occupant comfort, and cost indicators, were used to train an XGBoost ML model. The model achieved R² values above 0.97 and %root mean square error (%RMSE) values below 6.59% on the test data, demonstrating high predictive accuracy across all predicted outputs. The trained model was then embedded within a Streamlit-based interactive interface, enabling users to explore facade performance predictions in real time and support early-stage design decision-making without the need for additional EnergyPlus simulations.