The Relationship Between Streetscapes Around Tourist Attractions and Visitors’ Emotions from a Servicescape Perspective: Implications for Sustainable Tourism Based on Deep Learning

The external servicescapes surrounding tourist attractions influence visitor emotions and contribute to the sustainable development of urban tourism destinations. However, the relationship between objective streetscape visual elements and subjective emotions from a big data perspective remains unclear. This study examines their predictive capacity, relative model-based importance, and nonlinear association patterns using 14,968 street-view images from 178 tourist-attraction AOIs in Shanghai and UGC-derived emotion scores. Drawing on servicescape theory and a data-driven sustainability perspective, semantic segmentation was used to classify streetscape visual elements into three dimensions (environmental conditions, spatial layout and signage or symbols). Among the evaluated models, the MLP ensemble achieved the strongest out-of-sample performance, but its test-set R2 was only 0.026, indicating limited out-of-sample predictive capacity. SHAP analysis revealed that environmental conditions and spatial layout and functionality jointly accounted for 89.3% of the total absolute SHAP attribution, although the model’s overall predictive power remained low. PDP analysis further identified distinct nonlinear response patterns across streetscape features. This study extends servicescape research to open tourism environments and provides exploratory evidence for sustainable streetscape management around tourist attractions.