GuardianPath: Time-Aware Social Visibility Framework for Safe Pedestrian Route Navigation Using Machine Learning

GuardianPath presents a machine learning–driven framework for safe pedestrian route navigation in urban environments. The system operationalizes Jane Jacobs' "eyes on the street" theory by computing time-aware Social Visibility Scores using five engineered features—Proximity, Active Hour, POI Density, Anchor Presence, and Night Penalty—derived from OpenStreetMap data. An XGBoost regressor (R² = 0.9998) predicts segment-level safety, and a modified Dijkstra's algorithm generates safer routes. TreeSHAP provides per-route explainability. Evaluated on real-world Bengaluru road networks.