Spatio-temporal Similarity Measurement Of Inter-city Population Flow At Multi-scale Based On Migration Big Data

The population flows between geographic units,which are the comprehensive manifestation of logistics flow,information flow and other socio-economic connections,shaped our society and environment at multi spatial-temporal scales.The rapid development and popularization of ICT(information and communication technology)and LBS(location-based services)have boosted the capability for the continuous observation of the inter-city population flow on the one hand,while on the other hand,it also puts forward higher requirements for the expression and analysis of the characteristics of intercity population flow based on fine-grained big data.The characteristics inter-city population flow of an individual city often shows a multi-scale tendency which highly correlated with space-time.This tendency implies that the intercity population flow starting from a specific city is more inclined to take a specific set of cities as destinations at a specific time,which,from the perspective of data,can be reflected by multi factors like flow flowrate & direction,means of travel,and their temporal variation.However,the current complex network-based methodology of intercity population flow lacks detailed and dynamic expression & analysis of the characteristics of population flow of an individual city,and also,the spatial-temporal similarity between the tendencies of the inter-city population flow,which could be regarded as a comprehensive property of a city,has always been ignored.In such context,considering that the tendency of inter-city population flow in a single city is often manifested in the spatial-temporal variation of the flowrate,direction,travel mode and distance of the population flow,this paper introduces the characteristics of paradigm of similarity measurement in cognition science.Starting from the conceptual abstraction,definition,expression and similarity calculation of the characteristics of the inter-city population flow of a single city,it constructs the basic theory and time-space analysis methods for the measurement of the similarity of the inter-city population flow.On the one hand,it studies the basic theory of the similarity measurement of the spatial-temporal similarity of the inter-city population flow: Firstly,it studies the definition & expression of the population flow tendency(PFT).Also,the logical expression and physical calculation of population flow similarity index(PFSI)are explored to construct a unified comparison method between PFTs in multi time and space.On the other hand,the paper explores the multi-scale analysis method of spatialtemporal similarity between PFTs: Firstly,from the perspective of time,the paper studies the hierarchical clustering analysis,clustering result evaluation,semantic extraction & expression based on the temporal similarity matrix.Then,the time similarity analysis based on migration big data is carried out.Secondly,from the perspective of space,the hierarchical clustering analysis and clustering result evaluation method based on the spatial similarity matrix are studied.The local spatial autocorrelation analysis methods of PFTs are explored,and the case study of spatial similarity analysis is carried out based on the migration big data.The contribution and conclusion of the thesis in theory and knowledge discovery are as follows:(1)A new cognitive perspective and analysis method based on the similarity measurement of inter-city population flow of cities.Firstly,the thesis defined and expressed the concept of PFT considering multi factors(inter-city population flowrate,travel modes,distance,etc.)to character the inter-city population flow of individual city completely and dynamically.Secondly,faced with the complex,dynamic and multidimensional characteristics of PFT,this paper solved the logical expression and physical calculation problems of the similarity measurement between PFTs,and builds a highly interpretable similarity index of PFT,which provides multi-scale comparison methods between the characteristics of inter-city population flow of cities.(2)Multi-scale analysis strategies based on the spatial-temporal similarity measurement of inter-city population flow.The paper establishes multi-scale analysis strategies from the perspective of spatial/temporal similarity measurement of PFT,including hierarchical cluster analysis based on similarity matrix,evaluation of clustering results,semantic extraction & expression of clustering results.In addition,considering the spatial adjacency matrix of cities,the spatial similarity and attribute similarity of PFT are integrated to establish the global/local spatial autocorrelation index of inter-city population flow,which deeply explores the features of the PFT as a comprehensive property of individual cities.The proposed temporal analysis methods could be the substitution of the traditional empirical method to implement the multiscale feature identification and division of PFTs in the time dimension in a quantitative manner.The spatial analysis methods can deeply explore and unearth the similarity of the PFTs,which is an indirect relationship between cities that has been neglected by traditional research.(3)Reveal the multi-scale temporal and spatial characteristics of the intercity population flow in China from the perspective of similarity measurement.Firstly,based on the analysis of the comprehensive characteristics of the temporal/spatial autocorrelation matrix of the population mobility tendency,this thesis confirmed the temporal self-similarity and spatial autocorrelation characteristics of the inter-city population flow in China.Secondly,we explore various application scenarios of the constructed multi-dimensional similarity analysis methods.Based on the analysis of temporal similarity matrix,it is found that the temporal autocorrelation PFTs has scale differences dominated by geographical division and grades of city.While with hierarchical cluster analysis and spatial autocorrelation analysis of spatial similarity matrix,it is proved that the spatial autocorrelation characteristics of PFTs are also highly differentiated in multi scales dominated by geographical divisions and urban grades.In general,this thesis has implemented theoretical innovations in the definition,expression and similarity calculation of PFT,which is a detailed portray of the intercity population flow of individual city,and constructed a similarity index and analysis strategies of the PFT for better understanding,calculating,explaining and applications.It provides a new cognitive perspective for multi discipline to express and analyze the phenomenon of inter-city population flow,and has modest contribution to scientific methods and knowledge.