From points to patterns: An explorative POI network study on urban functional distribution

In the context of rapid urbanization, urban spaces not only accommodate a growing population but also produces complex socio-economic activities and cultural exchanges. Cities are complex systems, and conventional Points of Interest (POI) analysis methods, which usually assess the density and diversity of POIs in various neighbourhoods, often fails to capture this complexity. To address these limitations, this study introduces a novel approach by transforming POI sequences into words along streets and applying Latent Dirichlet Allocation (LDA) model to identify urban functional regions. Unlike traditional approaches that rely on subjective delineation of administrative boundaries, Voronoi cells or regular grids, our approach identifies street level functional areas that align more closely with human experience. Based on these functional topics, a multi-layered Poi-Topic network is then constructed to help better understand the roles specific POI plays within urban functional regions. This approach effectively distills the spatial distributional patterns of urban functions and provides a micro-level foundations for analysing the contextual interrelationships between POIs, thereby offering a more nuanced understanding of urban spaces. The effectiveness of the approach is demonstrated through the London case study. The results show that the proposed approach can effectively identify and delineate urban functional areas based on the co-occurrence patterns and network structure of POI vocabularies. The network centrality analysis further reveals the structural properties and interaction patterns, providing valuable insights into the roles and positions of different POI types in the functional organization of urban space. This method of using POI sequences and network analysis offers a new tool for urban planners, geospatial scientists, and policymakers, enabling them to understand and plan urban spaces with greater precision. • Introduces a novel street-level approach integrating topic modelling and network analysis for urban functional area identification. • Constructs multi-layer POI-Topic networks to reveal semantic and spatial structures of urban functions at the street scale. • Identifies six distinct urban functional topics in London using LDA on street-level POI sequences, providing granular insights into urban fabric. • Reveals varying network characteristics across functional topics, reflecting diverse urban organizational logics at human-experience scale. • Demonstrates the effectiveness of network metrics in uncovering key POIs shaping street-level urban functional structures. • Offers a human-centric perspective on urban functionality, supporting evidence-based planning and policy-making.