Identifying functional agglomerations and urban centers using open-source data and machine learning : framework, applications and planning implications

Metropolitan polycentricity is a topic that has been extensively studied. Traditional approaches to identifying urban centers have relied on high-resolution survey data, which is limited in many regions. More recent studies utilizing open-access point of interest (POI) data have primarily focused on POI density, often overlooking functional diversity and interactions. This paper introduces a theoretical framework for identifying distinct urban functional agglomerations and centers using open-source POI data. Considering both density and the combination of urban functions, it advances beyond previous density-centered methodologies by revealing multifunction features. Applying k-means clustering to Guangzhou, this study identifies 63 agglomerations and 11 centers, consistent with the city's Urban Development Plan (2018-2035). The results reveal the multifunctional nature of concentrations across multiple scales. The findings offer policy insights for improving land use, facility investment, and transportation planning. The proposed analytical framework can be readily applied to cities or countries with available open-source POI data.