Cultural heritage corridor networks in southern Anhui using machine learning and circuit theory

Traditional villages are important carriers of regional cultural heritage, yet urbanization and spatial fragmentation have weakened connections among them. This study develops a data-driven framework integrating spatial statistics, XGBoost–SHAP–Gain–AHP weighting, circuit theory, and network topology to construct a cultural heritage corridor network for 455 traditional villages in southern Anhui, China. Thirteen primary and ten secondary cultural source areas were identified by combining Getis–Ord Gi* hotspots with DBSCAN clustering. Protected cultural heritage site density received the highest resistance weight (0.382). The model identified 47 potential corridors with a total length of 2543.7 km and 603 km² of pinch-point areas. The network indices (α = 0.61, β = 2.04, γ = 0.75) indicate that the model-derived potential corridor network exhibits a certain degree of loop redundancy and a relatively high level of potential connectivity.