Spatial distribution patterns and influencing factors of traditional villages in southwest China: an integrated MGWR-XGBoost approach

Traditional villages in Southwest China exhibit complex spatial distribution patterns, yet comprehensive quantitative analysis of their clustering characteristics and influencing factors remains limited. This study integrates Multi-scale Geographically Weighted Regression (MGWR) with eXtreme Gradient Boosting-SHapley Additive exPlanations (XGBoost-SHAP) models to analyze spatial distribution patterns and influencing factors of 2,177 traditional villages across Southwest China. Results reveal: (1) significant hierarchical clustering with high-density concentrations in northwestern Yunnan and southeastern Guizhou, (2) pronounced spatial non-stationarity in factor influences across different geographical scales, (3) intangible cultural heritage density as the primary influencing factors with the highest SHAP importance value, and (4) complex nonlinear relationships and synergistic interactions among cultural, environmental, and accessibility factors. This integrated methodological framework advances heritage science by providing interpretable machine learning tools for spatial pattern analysis, offering scientific foundation for differentiated conservation strategies in mountainous regions.