Enhancing Urban Vitality through Big Data: A Case Study of Yinchuan City Using GWR and GBDT Models

Abstract Urban vitality, indicative of human activity in spaces, often eludes real-time analysis due to its complex nature. This research, with Yinchuan City as a case study, leverages Baidu heat maps as proxies for urban vitality and employs data from Baidu Street View, POI, and traffic within a 5D framework as independent variables. The study unfolds in phases, initially applying spatial analytics and deep learning to scrutinize built environment variables linked to urban vitality. It then uses ordinary least squares (OLS) to pinpoint influential factors and Moran’s I to assess the spatial autocorrelation of urban vitality. Geographically Weighted Regression (GWR) is employed to explore spatial heterogeneity, while Gradient Boosting Decision Tree (GBDT) analysis discerns variable importance for strategic planning. Results reveal a significant impact of built environment variables on Yinchuan’s urban vitality, with a noticeable positive autocorrelation and spatial clustering in central urban areas like Xingqing, Jinfeng, and Xixia districts. GWR analysis delineates a pattern of agglomeration in these central areas. Insights from the GBDT model inform priority-setting in planning. Recommendations offered by this study aim to elevate urban management and address urban challenges to enrich the living environment and invigorate urban vitality.