GeoUrbNet: A Multi-Source Graph-Based Framework and Dataset for Urban Building Function Analysis

This project presents a multi-source urban building function analysis framework integrating shapefile-based geometry, remote sensing imagery, street-view semantics, POIs, and population density. We construct a knowledge graph linking over 1 million buildings in Wuhan (2009–2021), generate a GraphSAGE-based prediction model, and propose the ST-GraphMiner algorithm for temporal pattern mining. All datasets (shapefile and graph database), prediction results, visualizations, model weights, and user documentation are included. This work has supported two journal manuscripts focused on framework methodology and dataset construction, respectively.