Predicting Lobby Location as an Indicator of Occupant Comfort: A Machine Learning (ML) Approach

This study investigates occupant comfort and spatial preference in hospital lobbies of Dhaka, Bangladesh, addressing the gap in data-driven analysis of hospital lobby spaces. Environmental, spatial, and behavioral data were collected and analyzed using three Machine Learning (ML) models: Decision tree (DT), Random Forest (RF), and Extreme Gradient Boosting (XGBoost). RF achieved the highest accuracy of 0.96 with RMSE of 0.65, identifying key factors affecting the lobby satisfaction. SHAP analysis showed that spatial attributes, especially Floor_Level and Window_State, are dominant over environmental conditions. This study shows the potential of ML to support Evi dence-Based (EB) user-centered design in urban hospitals.