Modeling and vitality coupling of urban and rural functional area based on POI data and GCNN

With the rapid accumulation of multi-source spatial data in cities, the identification of urban and rural functional zones and the modeling of dynamic structures have become important research directions in urban intelligent planning. However, existing methods are difficult to fully reflect the functional heterogeneity and dynamic vitality performance in complex regions. Therefore, this study proposes an urban–rural functional vitality recognition framework that integrates semantic modeling and graph neural networks. It integrates theme modeling and word vector construction to construct multi-semantic features and identifies functional regions through density clustering. At the same time, it constructs a graph structure based on spatial adjacency relationships and uses graph convolutional networks for functional category recognition. In performance testing, when the number of iterations reached 300, the F1-score of the proposed model was 0.91, the coupling correlation coefficient was 0.834, and the boundary F1-score was 0.82. In the actual application test within the Fifth Ring Road area of Beijing, the Olympic Park scored 0.813 on the leisure park, indicating a highly concentrated spatial response. The experiment shows that the proposed method achieves coupling modeling consistency in the area within the Fifth Ring Road of Beijing, and demonstrates good capabilities in urban function classification, spatial relationship modeling, and vitality response analysis. This study aims to provide intelligent support for urban–rural spatial governance and data-driven urban planning.