Social media interaction and built environment effects on urban walking experience: A machine learning analysis of Shanghai Citywalk

In fast-paced urban environments, Citywalk has emerged as a key leisure activity for urban residents to alleviate stress and enhance emotional well-being. From the perspective of virtual-physical interaction, this study integrates social media data with geospatial information, utilizing machine learning methods and spatial statistical analysis to explore the multidimensional driving mechanisms and complex relationships affecting the emotional experiences of Citywalk participants. The findings indicate that the interaction index, as a core indicator of virtual social behavior, plays a key role in influencing emotional scores (SHAP value = 4.9104), exhibiting progressive effects without evident threshold characteristics. POI density demonstrates significant nonlinear threshold effects, with marginal benefits substantially increasing when density reaches 44.06. Additionally, spatial autocorrelation analysis of emotional scores reveals spatial clustering patterns, underscoring the critical role of interactions between virtual social behavior and physical spatial elements in emotional generation. In comparison, functional diversity and transit accessibility exhibit weaker but complementary effects on emotional scores. This research quantifies the roles of digital social behavior and the built environment in shaping emotional experiences from a virtual-physical interaction perspective, uncovering how virtual social behavior integrates into social space production through individual perception and social interaction. It extends theoretical frameworks in social space production and emotional geography. The findings provide data-driven insights for optimizing urban walking space design, proposing interaction index-oriented strategies to promote synergy between virtual and physical spaces, thus facilitating the creation of high-quality, emotionally friendly urban environments.