Spatiotemporal evolution and influencing factors of regional resilience in China: evidence from provincial panel data using Spearman-CRITIC and XGBoost-CatBoost-SHAP

Abstract

Regional resilience is closely related to sustainable development because it reflects the capacity of a region to maintain stable economic, social, ecological, and infrastructural functions under long-term transformation. Based on provincial panel data for China from 2007 to 2024, this study constructs a regional resilience evaluation system covering economic resilience, social development resilience, ecological and environmental resilience, and infrastructure resilience. A Spearman-CRITIC model is used to calculate the resilience index, while cold–hot spot analysis and kernel density estimation are applied to examine spatial patterns and distributional dynamics. An XGBoost-CatBoost hybrid model and SHAP approach are further used to identify the explanatory contribution of potential influencing factors. The results show that China’s provincial resilience generally improved during the study period, although regional differences remained evident. Hot spots were mainly concentrated in eastern coastal and some central-eastern provinces, while cold spots appeared intermittently in western and southwestern areas. The SHAP results indicate that income level, private economic vitality, and urbanization rate make relatively large contributions to model prediction, with clear regional heterogeneity. Robustness tests using alternative weighting methods support the stability of the main explanatory patterns. These findings provide empirical evidence for understanding regional resilience improvement under sustainable development-oriented governance.