Revealing the Relationship Between Spatial Visibility and Personnel Flow in Hospital Outpatient Departments Using Graph Deep Learning and Agent-Based Modeling

This study integrates Graph Deep Learning (GDL) and Agent-Based Modeling (ABM) to analyze the association between spatial visibility and human flow in hospital outpatient departments. Addressing limitations of subjective traditional methods, spatial syntax theory is applied to construct a visibility graph model, quantifying node-level visibility. ABM simulates patient behaviors to capture environment-behavior interactions. Innovatively, Graph Neural Networks (GNNs) are employed to mine spatial network topology, embedding visibility indices and behavioral data into node-edge relationships to reveal underlying spatial-crowd dynamics. Trained on 100 outpatient samples, the model achieves 77.5% prediction accuracy. Results indicate positive correlations between visibility and spatial capacity demands. The research validates GNN’s interpretability in spatial analysis, offering a quantitative basis for optimizing hospital layouts and advancing evidence-based healthcare design. This framework bridges spatial metrics with behavioral patterns, enhancing data-driven decision-making in architectural planning.