Exploring spatiotemporal patterns of geosocial media data for urban functional zone identification

Abstract

Urban Functional Zones (UFZs) can be identified by measuring the spatiotemporal patterns of activities that occur within them. Geosocial media data possesses abundant spatial and temporal information for activity mining. Identifying UFZs from geosocial media data aids urban planning, infrastructure, resource allocation, and transportation modernization in the complex urban system. In this work, we proposed an integrated approach by combining the spatiotemporal clustering method with a machine learning classifier. The spatiotemporal clustering method was used to mine the spatiotemporal patterns of activities, of which the distinctive features were extracted as inputs into a machine learning classifier for UFZ identification. The results show that more than 80% of the UFZs can be correctly identified by our proposed method. It reveals that this work serves as a functional groundwork for future studies, facilitating the understanding of urban systems as well as promoting sustainable urban development.