Research on the spatiotemporal evolution and associated factors of seismic resilience in western China using machine learning

Western China faces significant seismic risks and has a relatively weak socioeconomic foundation, making systematic evaluation of its comprehensive seismic resilience strategically vital for regional sustainable development and national security. Existing studies show clear limitations in dynamic assessment and analysis of associated factors. This research constructs a comprehensive evaluation index system covering economic, population, infrastructure, and governance dimensions. Using the entropy weighting method and a weighted sum model, we measure the resilience index for 12 western Chinese provinces from 2000 to 2024, and apply Gaussian Kernel Density Estimation and machine learning methods to reveal the spatiotemporal evolution and economic explanatory factors of seismic resilience. Key findings include: (1) The mean regional resilience index increased significantly from 0.115 to 0.296, a rise of 157.4%, yet interprovincial disparities widened; (2) Resilience shows a spatial pattern characterized by higher levels in the southwest and lower levels in the northwest, with governance resilience receiving the highest entropy-based weight (0.473) and exhibiting the largest internal gap. A supplementary equal-weight sensitivity analysis confirms the stability of main results; (3) The Random Forest model achieves the highest predictive accuracy ( R 2 = 0.920) and identifies fixed asset investment and fiscal variables as important explanatory variables, with the Geodetector method further validating these findings. Based on these results, we propose differentiated policy implications for resilience-leading zones, key enhancement zones, and foundational strengthening zones, thereby offering practical references for improving seismic disaster prevention capabilities in Western China and advancing risk governance research in high-vulnerability regions.