Analysis On Urban Visitor Activity Patterns Based On Mobile Phone Data

The development vitality of a city depends on the ability to provide products and services for other cities and regions.When China’s urbanization has entered a stage of development with urban agglomerations as the main body,the communications between the central city and other cities in the urban agglomeration are increasing,and the radiation range of some public service facilities in the central city extends to the entire urban agglomeration.The spatial distribution estimation of resource allocation elasticity,the theoretical supplement of urban external space planning,and the scientific and precise decision-making of external traffic,are needed to be established on the basis of understanding activity patterns of urban visitors.Existing statistics are difficult to deal with the management challenges brought about by the massive scale and complex behavioral patterns of urban visitors.Traditional traffic models are based on aggregated data and cannot obtain the structural characteristics of individual travel chains.As a result,the requirements of service refining,monitoring normalization,and timely responses of public transportation in this new era cannot be adapted to.In recent years,with the popularization of smart mobile devices and the development of mobile Internet,mobile phone data have become low-cost technical means to comprehensively and accurately extract the behavioral characteristics of visitors on account of the characteristics of observation continuity and continuous tracking of behavior subjects.The research goal of this dissertation is to construct a systematic method for extracting visitor spatial activity information from mobile phone data,describing visitors in multiple perspectives such as group categories,spatial structure,and space-time distribution as far as possible,creating basic conditions for understanding its inherent laws and gradually establishing effective forecasting methods.Taking Shanghai as an example for empirical research and following the logic main line of "research foundation-functional support – pattern discovery – application of conclusions",this dissertation constructs "an analysis framework on visitor activity patterns based on mobile phone data" on the premise of consolidating the basis of visitor identification,to describe the multi-dimensional characteristics of visitors by referring to the theoretical achievements of related disciplines: extracting activity categories based on the rules of the interest value of visitor spatial activity point distribution,using the community detection algorithm of complex network to identify the spatial agglomeration structure of the city’s external space,and describing the spatial distribution characteristics and temporal changes of visitors based on the eigenvectors extracted by the matrix dimensionality reduction technology.The technical contribution of this dissertation is verified through practical applications of Shanghai public transportation service system construction.The main research work and findings can be summarized as the following four aspects:(1)Visitor identification and entrance identification method based on mobile phone dataThe accuracy of visitor identification is the basis of following research work.In this dissertation,given that there is only one single city and no label of user register location in the mobile phone data,four indicators are determined to distinguish the activity characteristics of visitors and residents(the number of active days,the number of signaling segments,whether there are nighttime stays,and the information entropy of rest places at night)and nine rules are proposed.As a result,54.9% of the total active mobile phone users are identified as visitors,and the daily average proportion of visitors to the total number of the population is 26.8%.This method improves the identification rate of visitors by 55% compared to the method of only using active days.Entrance is an important feature of visitor intercity travel.It plays an important role in the statistics of visitor "points of interest".This dissertation proposes a visitor entrance identification method based on mobile phone data.In the data preprocessing progress,the twodimensional data of base stations are converted into one-dimensional data of mileage to optimize the calculation accuracy of the features.By inferring railways and their stations with zone sequences where visitor signal trajectories pass in the railway network,five indicators(direction,distance,average speed,the validity of travel data,and the number of valid records)are constructed to distinguish railway travels,and seven rules are proposed for verification.This method can effectively solve the problem of railway station identification with missing signalling data in the base stations around the railway stations.The identification rate for visitor rail travel is 7.4 times higher than that achieved by the method of identifying entrances based only on the location of railway stations.The above methods provide basic information,such as the visitor label,signaling segment labels,the locations of rest places at night,moving/activity segments,and entrance,for subsequent researches.(2)Proposing a method of visitor activity category classification from the perspective of group classificationThis dissertation puts forward the concept of "point of interest" which eliminates the interference of stop points at terminals and rest places at night according to the characteristics of visitor activities.Considering the planning and purpose of visitor time allocation,two classification indicators,which are the dispersion degree and the core degree,are defined.The activity behaviors in space of visitors are summarized into five activity categories,namely the single point-of-interest structure(54.4%),the 4-7-secondary-point-of-interest single-chain structure(2.7%),and the 2-4-point-of-interest single-chain structure(9.6%),the two-point-ofinterest round-trip structure(32.8%),and the three-point-of-interest-level structure(0.5%).(3)Identifying the community structure of visitor activity space from the perspective of spatial structureThis dissertation uses visitor movement information in the basic spatial unit of the raster to model the network.The urban external space connection network is constructed,and the Louvain algorithm is used for community detection.The spatial expression form of the community is established in urban external space.It is found that there is a three-level structure,which is different from the hierarchy structure of the city overall planning.The larger the connection ratio of the communities is,the closer these communities are to the central city.However,the central communities are less stable than those peripheral ones.More than 86%of the marginal zones have a less-than-0.2 ratio which is the strength proportion of the connections between the marginal rasters and external communities to the connections between the marginal rasters and the local community.This indicates that there is sufficient credibility in this community structure.(4)Decomposing the basic patterns of visitor activity characteristics from the perspective of the essence of time and spaceThis dissertation uses the singular value decomposition method to decompose the spatiotemporal matrixes of high-dimensional community activity level and community connection strength from the perspectives of the whole and the different patterns.It is found that these matrixes can be simplified into 3 modes—daily stable change mode and holiday/working day local active mode.Using the multiple linear regression method,the contribution level of different activity categories to the overall spatiotemporal characteristics can be obtained.The categories of visitors that should be focused on in policy making are found(the single point-of-interest structure,the 4-7-secondary-point-of-interest single-chain structure,and the three-point-of-interest-level structure).The fitting degrees of these models are all higher than 0.94.This dissertation presents the decision support process that converts data resources into decision-making capabilities through two application scenarios of hub transportation and tourism transportation.In terms of hub transportation,by comparing the differences in the distribution of visitor activity characteristics and spatial correlation distribution between different entrances,it is demonstrated that the hub transportation system needs to consider the visitor activity characteristics for optimization.It is found that the service area and effectiveness of Shanghai Railway Station public transportation is the largest,Hongqiao Terminal public transportation comes the second,and Shanghai South Railway Station and Pudong Airport public transportation rank at last.New bus lines are needed to be added between Putuo District and Shanghai Hongqiao Railway Station.In terms of tourism transportation,by calculating the correlation strength between tourism facilities,it is found that tourists’ travel purposes show a trend of diversification;Dongping National Forest Park,Happy Valley,National Convention and Exhibition Center,Shanghai Science and Technology Museum,Shanghai Zoo,Wildlife Park,etc.need to build tourist bus lines between accommodation and tourist attractions.The priority to adjust or establish bus routes can refer to the improved method of accessibility that comprehensively considers supply,demand and economic benefits.It is recommended to develop a multi-level tourist bus network.Based on the urban visitor activity pattern in the big data environment,this dissertation discusses and analyzes visitor activity characteristic description,and its results reveal the visitor activity laws,improving the ability of planners and decision makers to accurately grasp the development trend of transportation,providing broad application prospects for the refined and quantitative strategy optimization of urban public transportation service system,and guiding the comprehensive integration research of urban and regional space.