Abstract: Understanding the relationship between housing prices and the surrounding environment is essential for optimizing spatial resource allocation and advancing refined urban governance.Despite its importance,the influence of visibility attributes on housing prices remains insufficiently explored.This study integrates four 3D visibility metrics into a machine learning-enhanced hedonic pricing model (HPM) and employs SHAP(shapley additive explanations) analysis to elucidate the heterogeneous effects of these metrics on housing prices in Shenzhen.The enhanced HPM substantially improves prediction accuracy,with the random forest model outperforming ordinary least squares and XGBoost approaches,achieving the highest explanatory power(R²=0.881).Visibility metrics account for 38.4% of the explained variance in housing prices.The analysis reveals that a higher visible volume of sky exerts a suppressive effect on housing prices,whereas longer average sightlines,larger visible facade areas,and greater visible volume of buildings are positively associated with price increases.These findings offer novel insights into the visual dimension of urban form and its economic implications,contributing to human-centered and sustainable urban development strategies.

Key words: visibility analysis, hedonic price model, SHAP analysis, random forest, 3D buildings