What Makes Visitors Want to Stay? An Explainable ML Study of Panoramic Visual Environments and Momentary Affect

Visitors’ intention to stay at different locations within urban parks may be associated with visible surroundings and momentary affect. At 60 fixed sites in Anhui Garden of Hefei Garden Expo Park, we captured panoramic images on both survey days (120 images in total) and obtained 600 valid on-site questionnaires. Visual composition was quantified using semantic segmentation. Spearman correlations, fivefold site-held-out cross-validation, and TreeSHAP analysis were then used to examine site-level associations and model predictions. In the analysis of two-day averages across 60 sites, positive and negative affect correlated with mean intention to stay in opposite directions (ρ = 0.444 and −0.440, respectively). Perceived variety showed the strongest correlation (ρ = 0.824). Vegetation pixel share was negatively correlated with intention to stay (ρ = −0.426), whereas sky, water, and built-interface shares were positively correlated. Ridge regression and XGBoost models using five visual features had out-of-fold mean absolute errors (MAEs) of 0.481 and 0.486 points, compared with 0.543 for the mean baseline. Adding concurrent affect reduced MAE to 0.391 for ridge regression and 0.388 for XGBoost, although the gains were inconsistent across survey days. In the visual-only XGBoost model, water, vegetation, and built interfaces had the highest mean absolute TreeSHAP values. At the two-day aggregated site scale in Anhui Garden, concurrent affect provided additional information for predicting mean intention to stay beyond panoramic visual composition.

Authors

Institutions

Publication Details

Journal
Land
Published
2026-11-16
DOI
https://doi.org/10.3390/land15112192
Primary Topic
Urban Green Space and Health
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

What Makes Visitors Want to Stay? An Explainable ML Study of Panoramic Visual Environments and Momentary Affect

Le Zhang, Qiang Wang, Chengyuan Cui, Yuting Tang
5 links
Land
Urban Green Space and Health
article

What Makes Visitors Want to Stay? An Explainable ML Study of Panoramic Visual Environments and Momentary Affect

Le Zhang, Qiang Wang, Chengyuan Cui, Yuting Tang
article en

Abstract

Visitors’ intention to stay at different locations within urban parks may be associated with visible surroundings and momentary affect. At 60 fixed sites in Anhui Garden of Hefei Garden Expo Park, we captured panoramic images on both survey days (120 images in total) and obtained 600 valid on-site questionnaires. Visual composition was quantified using semantic segmentation. Spearman correlations, fivefold site-held-out cross-validation, and TreeSHAP analysis were then used to examine site-level associations and model predictions. In the analysis of two-day averages across 60 sites, positive and negative affect correlated with mean intention to stay in opposite directions (ρ = 0.444 and −0.440, respectively). Perceived variety showed the strongest correlation (ρ = 0.824). Vegetation pixel share was negatively correlated with intention to stay (ρ = −0.426), whereas sky, water, and built-interface shares were positively correlated. Ridge regression and XGBoost models using five visual features had out-of-fold mean absolute errors (MAEs) of 0.481 and 0.486 points, compared with 0.543 for the mean baseline. Adding concurrent affect reduced MAE to 0.391 for ridge regression and 0.388 for XGBoost, although the gains were inconsistent across survey days. In the visual-only XGBoost model, water, vegetation, and built interfaces had the highest mean absolute TreeSHAP values. At the two-day aggregated site scale in Anhui Garden, concurrent affect provided additional information for predicting mean intention to stay beyond panoramic visual composition.

LandVol. 15(11)
Anhui Jianzhu University (CN)
Sustainable cities and communities
Urban Green Space and Health
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon