Spatial Heterogeneity of Built Environment Perception's Impact on Street Vitality: A Multi-Level Interpretative Framework Based on MGWR-SHAP

Street vitality is crucial for sustainable urban development, yet current understanding of how built environment perceptions influence vitality remains limited by global statistical approaches and the lack of interpretable frameworks for analyzing spatial heterogeneity. This study proposes a novel multi-level interpretative framework combining Multiscale Geographically Weighted Regression (MGWR) with SHAP values to examine spatial variations in perception-vitality relationships. Using multi-source data from Hohhot, China, including mobile phone signals, POI data, and street view imagery, we analyzed how four dimensions of environmental perception influence street vitality across different urban contexts. The analysis reveals significant spatial heterogeneity in perception-vitality relationships, with varying effects across urban locations. Pleasure perception shows the strongest positive influence (SHAP values 0.072-0.103), while convenience perception exhibits an unexpected inverse U-shaped relationship with vitality. The machine learning approach (R² = 0.421) outperforms traditional methods in capturing nonlinear effects and complex interactions. The findings demonstrate the importance of considering both spatial heterogeneity and nonlinear relationships in understanding street vitality, suggesting the need for context-sensitive approaches to urban design and planning interventions.