Quantify city-level dynamic functions across China using social media and POIs data

Location-aware big data from social media have been widely used to study functions of different zones in a city but not across a city as a whole. In this study, a novel framework is proposed to quantify city-level dynamic functions of 200 cities in China from a perspective of collective human activities. The random forest model was used to determine the temporal variations in the proportions of different urban functions by examining the relationship between Points-of-Interest (POIs) and Tencent Location Request (TLR) data. We then hierarchically clustered and analyzed the structures and distribution patterns of the dynamic urban functions of 200 Chinese cities at different temporal scales. In the end, we calculated an urban functional equilibrium index based on the urban functional proportion and then mapped spatial distribution patterns of the indexes across mainland China. Results show that on a daily scale when the cities were grouped into two clusters, they are either dominated by the work/education and commerce or residence functions. The cities in the former cluster are mainly the provincial capitals and located within major urban agglomerations. When the cities were grouped into four clusters, the clusters are dominated their commerce, work, residence, and balanced multiple functions, respectively. For each of the 200 cities, its urban functions change dynamically from the daybreak to the evening in terms of human activities. Besides, the equilibrium indexes show a power-law relationship with their rankings. Our research shows that city-level dynamic function can be quantified from the perspective of variations in human activities by using social media big data that otherwise could not be achieved in the conventional urban functions' studies.