Explorative Analysis of Wuhan Intra-Urban Human Mobility Using Social Media Check-In Data

Lin Li, Lei Yang, Haihong Zhu, Rongrong Dai, Claudia Torres Codeço
2015 PLoS ONE  
Social media check-in data as a geo-tagged information source have been used for revealing spatio-temporal patterns in the field of social and urban study, such as human behavior or public issues. This paper investigates a case study and presents a new method of representing the mobility of people within a city from check-in data. By dividing the data in a temporal sequence, this study examines the overall mobility in the case study city through the gradient/difference of population density
more » ... a series of time after computing the population density from the check-in data using an incorporated Thiessen polygon method. By classifying check-in data with their geo-tags into several groups according to travel purposes, and partitioning the data according to administrative district boundaries, various moving patterns for those travel purposes in those administrative districts are identified by scrutinizing a series of spatial geometries of a weighted standard deviational ellipse (WSDE). Through deep analyses of those data by the adopted approaches, the general pattern of mobility in the case city, such as people moving to the central urban area from the suburb from 4 am to 8 am, is ascertained, and different characteristics of movement in those districts are also depicted. Furthermore, it can tell that in which district less movement is likely for a certain purpose (e.g., for dinner or entertainment). This study has demonstrated the availability of the proposed methodology and check-in data for investigating intra-urban human mobility. In general, modeling human movement in the inter-city or intra-city is tightly associated with urban models or is a type of urban modeling itself. As early as 1946, G. K. Zipf [7] proposed the gravity model for simulating inter-city human mobility. Because the model contains undetermined parameters such as the evaluation indicators of cities and the travel distance defined in the damping function, it could not be applied when travel survey data are lacking. To solve this problem, F. Simini [8] proposed the radiation model, which is parameter-free and has a good performance in inter-urban human mobility prediction. The rank-based movement model [9] , the generalized potential model [10] , and the intervening opportunities model [11] are additional models used to simulate and predict human mobility patterns on different levels. Intra-urban human mobility has shown its great importance in traffic accessibility, locationbased services, and urban planning [12] [13] [14] , and has been explored by many studies for its regularity and predictability [2, 8, 15, 16] . These models all operate from multiple perspectives and take into account such factors as geography, population, and social economy. Some trends and intensities of human mobility are revealed by means of mathematical deduction from empirical data sets. Due to the growing popularity of location-based services and geo-social networks, users communicate more and more private location traces to service providers, as well as explicit spatio-temporal data, often called "check-ins", about their presence in specific spots or venues at given times [17] . Information about time and location as well as photos is included in a check-in record. We can acquire when and where the check-in activities happened and extract the footprints of a large number of individuals [1] . Check-in data as geo-tagged information sources has been used for revealing some spatio-temporal regularity in urban areas such as identifying commercial centers [18] , detecting local events [19] and determining population distribution [20] . For mining patterns of human mobility from check-in data, some studies substantially address that topic. Liu, Y. et al. [1] extract inter-urban movements in China from a check-in data set to analyze the underlying patterns of trips and spatial interactions and construct a spatial network where the edge weights denote the interaction strengths. They find that the communities detected from the network are spatially cohesive and roughly consistent with province boundaries. The approach links patterns at the collective level of spatial interactions versus the individual level of human movements from mobile phone or taxi data sets. Noulas, A. et al. [9] study urban mobility patterns in several metropolitan cities, which verify variations in human movement caused by different distributions of places across different urban environments, and indicate that the probability of transiting from one place to another is inversely proportional to a power of their rank. Wu, L. et al. [21] construct a temporal transition probability matrix to represent the transition probability of travel demands during a time interval, and adopt the mechanism of an agent-based model, which combines activity-based analysis with a movement-based approach. Most existing human mobility models capture real human movement to some degree, but assume continuous and homogeneous space [22] . The gap between assumption and reality could be bridged by some devised spatialization approach in processing check-in data. By far, all studies of human mobility based on check-in data are a type of case study revealing spatio-temporal features of human mobility. The intra-urban movement of individuals is affected by a number of factors, such as urban form, mode of transportation, transportation networks and socio-economic status [23] [24] [25] [26] . Movement patterns dependent on a specific city could not be applied to other cities or urban areas. Thus, the uniqueness of human mobility to the city requires investigating a city individually, and the revealed patterns or features are irreplaceable for the city in urban planning and decision-making.
doi:10.1371/journal.pone.0135286 pmid:26288273 pmcid:PMC4545943 fatcat:z4od34nz3bb3hbpszg2z3w5slu