Fig 1.
Research framework for analyzing virtual-physical interaction in Citywalk activities.
Fig 2.
Spatial distribution of key indicators for Citywalk activity in Shanghai (a) POI density (points per grid); (b) functional diversity (Shannon index); (c) transportation accessibility; (d) interaction index.
The base map data were obtained from Natural Earth (https://www.naturalearthdata.com/).
Fig 3.
Spatial distribution of Citywalk sentiment scores across Shanghai’s urban districts.
Base map source: Natural Earth.
Table 1.
Model comparison results.
Fig 4.
Model performance evaluation and feature importance analysis: (a) Taylor diagram comparing predictive performance of neural network (●), random forest (□), and XGBoost (△) models; (b) SHAP summary plot showing magnitude and distribution of feature impacts on sentiment score predictions; (c-f) Dependence relationships; (g) Force plot analysis.
Table 2.
Piecewise regression results.
Table 3.
Feature interaction analysis results.
Fig 5.
Nonlinear relationships between environmental features and sentiment scores: (a) Piecewise regression scatter plots showing threshold effects; (b) Correlation heatmap of environmental features.
Table 4.
Threshold analysis and connectivity evaluation results.
Fig 6.
Threshold analysis of the spatial weight matrix.
(A) Comparison of threshold distributions; (B) Connectivity analysis; (C) Spatial connectivity pattern of the grid diagonal method.
Fig 7.
Results of Local Moran’s I analysis showing spatial clustering patterns of sentiment scores in Shanghai.
Red areas indicate high-high clusters, blue areas represent low-low clusters, and light colors show spatially insignificant areas. Base map source: Natural Earth.
Table 5.
Spatial morphological characteristics of sentiment CLUSTERS.