Understanding urban community structure at multiple levels through crowd mobility

Abstract

Current research on urban community structure primarily focuses on a single level and often overlooks the associations among multiple regions within a city when detecting community structures. In this regard, this research employs a hypergraph model—capable of capturing multi-region relationships—and applies a hierarchical clustering approach to reveal a multi-level urban community structure. We conduct an empirical study of Shanghai's multi-level community structure using cell phone signaling and land use data, demonstrating that the proposed method effectively detects communities. Our findings indicate that: 1) Compared to graphs, hypergraphs more effectively represent complex inter-regional relationships, thereby improving the effectiveness of community detection; 2) Shanghai exhibits a structural pattern of multifunctional areas radiating from the city center, characterized by a high degree of land-use mixing; 3) Discrepancies exist between community structures derived from real crowd flow data and those envisioned in urban planning schemes, with certain communities merging or splitting. The model presented in this study can monitor current urban community structures and, when compared with established urban planning, guide more rational urban development. Additionally, it provides a method to support multi-relationship, multi-level studies of phenomena such as the urban heat island effect and traffic congestion.