Acta Geographica Sinica ›› 2016, Vol. 71 ›› Issue (3): 471-483.doi: 10.11821/dlxb201603010

• Geographic Information Analysis • Previous Articles     Next Articles

Discovering urban functional regions using latent semantic information: Spatiotemporal data mining of floating cars GPS data of Guangzhou

Shili CHEN1,2,3(), Haiyan TAO1,2(), Xuliang LI1,2, Li ZHUO1,2   

  1. 1. Center of Integrated Geographic Information Analysis, School of Geography and Planning, Sun Yat-sen University, Guangzhou 510275, China
    2. Guangdong Provincial Key Laboratory of Urbanization and Geo-simulation, Guangzhou 510275, China
    3. Urbanization Institute of Sun Yat-sen University, Guangzhou 510275, China
  • Received:2015-07-30 Revised:2015-11-27 Online:2016-03-25 Published:2016-03-25
  • Supported by:
    Projects of National High-tech Research, No.2013AA122302;Natural Science Foundation of Guangdong Province, No.S2013010012554;National Natural Science Foundation of China, No.41371499, No.41271138


China has been experiencing rapid urbanization at an unprecedented rate and as a result, urban internal space structure has evolved significantly. It is of great significance to label different functional regions (DFR) inside a city for urban structure analysis, policy making, and resource allocation. These DFRs include residential district, industrial district, education district, and the administration district. This paper explored the characteristics and distribution of urban functional regions based on big geographic data. With the latest road network data, the study area (i.e., 6 districts of Guangzhou city in Guangdong Province, China) was partitioned into 439 segments. By applying the employment of spatial and temporal semantic mining method to the one-week massive floating cars GPS data and the point of interest data, we developed a Latent Dirichlet Allocation (LDA) and Dirichlet Multinomial Regression (DMR) model. Moreover, OPTICS clustering method was employed to process the results of LDA and DMR to identify different functional zones. Meanwhile, status map of Guangzhou urban planning, and resident travel characteristics were used to verify the verification of mentioned results. The results show that this method can identify the obvious characteristics of urban functional areas, such as mature residential area, science and education culture area, commercial area, and development zone. The results also show that residential and commercial areas are dominant DFRs in Guangzhou city, which are surrounded by other types of functional regions. This paper brings a new perspective on using large-scale and high quality individual space-time data to study human migration and daily activities, as well as to explore social space to unveil the formation and mechanism of urban functional zones.

Key words: latent dirichlet allocation, functional regions, big geographic data, GPS data, point of interest, Guangzhou