Abstract:

Research on "city image" can facilitate urban culture perception, urban management and planning, and tourism resource development. In recent years, as intelligent mobile terminals and social media apps became increasingly popular, a large number of social media geo-tagged data containing text and location information have been generated, providing a new solution for city image perception studies. This article uses the social media geo-tagged data (Sina weibo check-in data in Beijing, 2016) to explore regional hotspots through spatial clustering, and mining the topics of different hotspots through two typical methods— term frequency-inverse document frequency (TF-IDF) and latent Dirichlet allocation (LDA). The results reflect the topics that users were concerned about and discussed in different places, revealing the culture, functions, and characteristics of diverse places of Beijing in great depth. The proposed city image abstraction approach by integrating text mining and spatiotemporal big data analysis can promptly expose the differences on themes of activities, attitudes, and preferences in different places in Beijing, thus reveal the social and cultural characteristics of the city. Our method is an important complement to the five-element model of city image, which focuses on the urban material form. In addition, the case study results of Beijing regional hotspots facilitate the preservation of city characteristics and shaping of space quality.

Key words: geospatial data, social media, microblog data, text mining, regional hotspot, city image