Street Vitality in Historic Cultural Districts Based on Multimodal Data and Explainable Machine Learning: A Case Study of Zhengding Ancient City

Historic cultural districts must balance heritage conservation with contemporary urban vitality, because excessive commercialization and overly static preservation can weaken both everyday use and cultural authenticity. This study examines 350 street segments in Zhengding Ancient City, China, by integrating street-view imagery, road networks, points of interest, population heatmaps, social-media text and images, historic-building records, and field surveys. Twenty-five indicators were organized into two independent axes: 11 environmental and spatial indicators formed spatial quality (Q), while 14 behavioral, perceptual, and cultural-communication indicators formed the Historic Street Vitality Index (HSVI). Random Forest, GBDT, XGBoost, and LightGBM were compared, and SHAP was used to interpret nonlinear associations. Vitality followed a pronounced main-axis–node clustering pattern, with several quality–vitality mismatches. Heritage concentration and historic–cultural visibility showed the strongest associations with HSVI, followed by functional provision and walking conditions; several factors displayed saturation or inverted-U-shaped relationships. Social-media evidence highlighted both cultural value and concerns about accessibility, services, traffic order, and commercialization. Six variables (E2, E3, F1, F2, S1, and S2) accounted for 71.43% of cumulative SHAP importance and can support routine screening, while the full Q–HSVI system remains appropriate for comprehensive diagnosis and periodic recalibration.