Assessing the visual walkability perceptions of urban streets via the lens of older adults: A case study in Shanghai

With the acceleration of population aging and urbanization, developing age-friendly cities has become a critical strategy to foster inclusive spaces for older adults. Walking plays a vital role in promoting health among older adults, which underscores the need for large-scale, cost-effective assessments of walking environments to guide urban street renovations. Based on a street view image dataset capturing the unique visual walkability perceptions (VWPs) of older adults, we developed a novel quantitative evaluation model using machine learning algorithms, and employed the SHapley Additive exPlanations (SHAP) method for model interpretation. The best-performing model was further applied to predict VWP scores and generate corresponding maps of Shanghai's central urban area. Additionally, comprehensive analyses have been conducted from multiple perspectives. The results show that the random forest algorithm performed well in predicting VWPs. Furthermore, vegetation, sidewalks, roads, and terrain are key street elements that positively influence VWPs. Preferences for these street elements vary across groups based on age, gender, and location. Meanwhile, the spatial distribution of VWP scores demonstrates significant scale effects, with spatial heterogeneity diminishing as the geographical unit size increases. Moreover, the geographical clusters of VWP scores have revealed the pronounced spatial inequalities in the development of age-friendly walking environments across the central urban area of Shanghai.