Understanding the Spatial Differentiation and Driving Mechanisms of Human Settlement Satisfaction Using Geographically Explainable Machine Learning: A Case Study of Xiamen’s Urban Physical Examination

In recent years, as Chinese cities have entered a stage of high-quality transformation, enhancing livability and achieving refined governance within existing urban spaces has become a central issue in urban planning and management. The establishment of the Urban Physical Examination mechanism has provided a scientific framework for evaluating urban performance. However, most existing studies focus primarily on objective indicators, paying insufficient attention to residents’ subjective perceptions and their spatial variations. As a result, the multi-scale mechanisms underlying human settlement satisfaction remain poorly understood. Using Xiamen City as a case, this study draws on data from the 2025 Urban Physical Examination Resident Survey and constructs a Geographically Random Forest (GRF) model to examine how block, community, housing, and personal attributes jointly shape human settlement satisfaction (HSS) and its spatial heterogeneity. The results show that (1) overall, block’ business vitality is the most influential factor affecting HSS, followed by community management and housing safety, highlighting the dominant roles of the built environment and grassroots management in shaping residential experience; (2) management and safety issues at the community level are more prominent in suburban areas, old neighborhoods, and zones surrounding tourist attractions, reflecting a mismatch between service provision and urban expansion; (3) housing-scale factors display significant spatial variation, with tenure and housing affordability emerging as key determinants of satisfaction among residents in newly developed districts; and (4) at the personal characteristic, age, residential duration, occupational prestige, and household income exhibit marked spatial heterogeneity, revealing satisfaction patterns jointly shaped by social mobility and urban growth. The study concludes that multi-scale spatial identification and resident perception feedback mechanisms should be strengthened within the Urban Physical Examination framework. Such efforts can promote a shift from static indicator monitoring to dynamic spatial governance, providing theoretical and methodological support for refined urban management and the improvement of human settlement environments.