Anna E. Avdiushina, Yulia A. Koroleva, Tatyana A. Markina, Igor A. Bessmertny

ITMO University

The article focuses on identifying informal urban areas based on data from news sources and social networks, utilizing geographical proximity as a criterion. A methodology for extracting geodata from texts for spatial clustering is proposed. Geographic names extracted from texts are transformed into geolocations through geocoding. The identified geopoints are then clustered by density, and a distribution of themes is determined for each cluster. This approach allows for an abstraction from administrative divisions to reveal clusters that are closer to the citizens' perception. The clustering results are promising for application in various urban infrastructure management tasks: monitoring public life, analyzing the quality of the urban environment, and public safety. The distinction of the proposed methodology lies in the synthesis of geodata for grouping objects. The software tools developed based on this methodology enable decision-making in the field of urban planning, including the development of city districts and transport infrastructure, the placement of socially significant objects, and ensuring safety.

data mining, smart city, information model, decision support systems, clustering, geospatial data