Nonlinear effects of the built environment on urban vitality in Jinan based on multi-source data and explainable AI

Understanding the nonlinear relationship between the built environment and urban vitality is a key challenge in urban planning. This study proposes an integrated analytical framework: (1) Combining multi-source data with explainable AI (XGBoost-SHAP) to decode nonlinear relationships and threshold effects; (2) Using the CRITIC weighting method for the first time to construct a composite urban vitality index; (3) Establishing a two-dimensional built environment framework using Baidu heat maps, nighttime light data, POIs, and street-view imagery to integrate "objective form" and "subjective perception". The findings reveal: (1) Urban vitality shows a distinct "east-strong, west-weak" clustering pattern; (2) Objective factors dominate, especially POI blending degree, density and spatial integration, while perceptual factors exhibit limited influence; (3) Key nonlinear thresholds are identified: POI blending best promotes vitality above 1.5; density shows diminishing returns beyond 4,000 units/km²; and openness follows an inverted U-shaped curve. This study promotes the methodology for measuring comprehensive urban vitality, enriches the dimensions of the traditional framework of built environment indicators, and provides a new analytical framework for planning in historically layered cities.