From Perception to Prediction: Modeling Pedestrian Satisfaction Using Multilevel Statistical and Sensitivity Methods

Abstract

This study presents an integrated modeling approach to evaluate pedestrian satisfaction in new urban cities characterized by rapid growth and limited multimodal connectivity. A structured questionnaire, distributed to stratified participants across residential, administrative, and service zones, captured user perceptions of 13 key urban design features, including safety, accessibility, visual coherence, and economic vibrancy. Descriptive statistics and visual analytics revealed that accessibility, protection from crime and traffic, and urban aesthetics were strong correlates of satisfaction. To model these relationships quantitatively, the study employed both ordinal and multinomial logistic regression, with the latter achieving 92.45% classification accuracy. K-means clustering and principal component analysis further uncovered latent user typologies, highlighting the heterogeneity of pedestrian priorities. Local and global sensitivity analyses, including mutual information metrics, identified easy access, protection from traffic, and crime prevention as the most influential features. Response surface modeling illustrated nonlinear interactions among key variables, emphasizing the multidimensional and synergistic nature of satisfaction outcomes. The findings showed that pedestrian experience is shaped not by isolated design features, but by their interactive effects across spatial, psychological, and infrastructural domains. The study offers actionable insights for human-centered urban design, while the presented analytical framework is scalable and supports evidence-based interventions in emerging urban contexts.