Shrinking cities, unshrinkable health: A study on the evolution of street spatial health performance in shrinking cities based on deep learning methods

The urban built environment has been proven to affect the health of residents, but the progress of research on urban spatial health performance (SHP) ² 2 The material space environment of the street can meet the public’s health needs and guide the conditions and degree of public health living. is constrained by lacking of material spatial data in geographical regions, and a few studies on urban SHP lack consideration of characteristics in population-shrinking areas. This study used Partial Least Squares Structural Equation Modeling (PLS-SEM) to construct a scientific evaluation system for urban street SHP. We selected eight shrinking cities in China for spatiotemporal distribution analysis of street SHP, used Baidu Street View (BSV) images and machine learning to automatically describe the characteristics of street spatial environment in shrinking cities. The health performance index growth rate GR_ISHP and Lasso regression model were adopted to analyze the evolution patterns and influencing factors of street SHP. Research has found that the street SHP of shrinking cities did not shrink completely, and there were similar factors influencing the street SHP in shrinking cities with different population sizes and geographical locations. Therefore, based on the above conclusions, this study proposed to classify urban streets into consolidated, improved, and maintained types, and respectively put forward optimization strategies. Due to the universality of the evaluation system and the accessibility of Baidu Street View, the research methodology can be reused in other cities in China. The findings of this research will contribute to optimizing the health performance of urban street space and urban management from the perspective of material spatial elements of urban streets.