Modeling the Street Staying Willingness Evaluation and Influence Factors Analysis Based on Street View Images

Street stayability is recognized as an essential dimension for assessing street vitality. Previous studies have often struggled to balance large-scale and micro-scale analyses. In this study, the concept of “Street Staying Willingness” (SSW) is introduced as a subjective indicator for street stayability assessment. The main objective is to conduct a large-scale assessment of street stayability based on street view images (SVIs) through machine learning to mimic human perception. Shapley additive explanations (SHAP) are employed to explore the complex relationships between SSW and street visual elements. The framework’s effectiveness as a substitute for field research was validated. The findings indicated that six street visual elements, namely terrain, roads, fences, walls, pedestrians, and sidewalks, exerted the most significant influence on SSW. Among these, pedestrians exhibited a notably pronounced linear relationship, while no particularly distinct linear relationships were observed between other visual elements and SSW.