Study On The Modeling And Evaluation Of Urban Spatial Interactions Based On The Mobile Phone Location Data

Urban space has always been the extremely important place for various human activities.Nowadays,with the continually increasing of urban population and sprawl of urban space,it is of critical challenge and urgent need to qualitatively understand and quantitatively analyze the complex interaction and relationship between massive human mobility and limited urban space,as well as to conduct a reliable prediction of the spatialtemporal travel patterns and distributions of citizens.Thus,forward-looking evaluations of the adaptability between human activities and urban spatial structures could be investigated to help solve the various problems emerged in the rapid process of urbanization.Fortunately,the fast development of information and communication technology,along with the location aware technology have aided the acquisition of various kinds of big data in a more convenient and real-time way,such as geo-tagged social media check-in data,mobile phone data,smart card data,and so on.These data could effectively capture the spatialtemporal footprints of large amount of citizens,and contains tremendous information regarding human activities,which brings many new insights into the study of human dynamics and spatial interactions.Among all the geo-tagged big datasets,the mobile phone location data,generated by the mobile phone users when communicating with the cell phone towers,is very special data because mobile phones have an extremely high penetration rate,and people usually take their cell phones with them.Mobile phone location data has the incomparable advantages of adequately wide sampling in human groups,time and space,thus,lots of researchers view this kind of big data as a reasonable source to describe human mobility.Therefore,this dissertation aims at making full use of the spatial and temporal sampling characteristics implanted in the mobile phone location data to explore human mobility patterns and model the spatial interactions from both qualitative and quantitative perspectives.Overall,this paper quantitatively investigates the biases of sparsely sampled location data in characterizing human mobility patterns,and proposes a method to select the more effective sampling locations from massive mobile phone towers for high-accuracy spatial interaction model calibration,as well as gives a new insight into the modeling of nonstationarity inhered in the spatial interactions under the complex and heterogeneous urban environment.More specifically,the contents of this paper are fourfold:1.Quantitatively investigating the representativeness and bias of sparsely sampled mobile phone location data in human mobility research.Mobile phone data is featured from traditional tracking data because the location records of each individual are quite sparsely sampled from both temporal and spatial perspectives.The biases inhered in this data should be addressed before using it to derive human mobility patterns.Thus,this paper proposes a linear regression model to quantify the aggregated underestimation level of given time segments in depicting human mobility indicators.Furthermore,the general law of bias based on different time segments is revealed,which could be a useful guidance for selecting reasonable dataset for human mobility and spatial interaction research.2.Exploring the effects of sampling locations for calibrating the globally spatial interaction model.As have been illustrated before,mobile phone location data has the advantage of widely sampling in urban space,however,"are the whole sampling locations always the better solution for spatial interaction model calibration" needs to be revealed in the era of big data.Moreover,this paper proposes a method to select the more effective sampling locations from massive mobile phone towers for high-accuracy parameters calibration,which could contribute to the better big solutions for spatial interaction model calibration,as well as the selections of surveying locations.3.Evaluating and modeling the nonstationarity existed in the spatial interaction between human activities and the heterogeneous urban space.According to the Toble's First Law of Geography,this paper assumes that close citizens present close preferences in space,thus,a locally spatial interaction model based on the spatial nonstationarity is proposed and also calibrated by the geographically weighted regression model.Finally,we analyze the spatial distributions of locally calibrated parameters combined with the topographic and geomorphic conditions of study area.The proposed method could be a useful guidance to the realization of fine spatial interaction modeling.4.Exploring the general evolution rules of scale effect on the spatial interaction modeling and evaluation.The scale effect is an important part of the classical modifiable areal unit problem hidden in the spatial analysis of dynamic data.This paper divides the urban space into different sizes of regular grid cells,and visualizes the aggregated interaction flows under different spatial scales cooperated with a comprehensive analysis of the variances.Then,both of the globally and locally spatial interaction models are calibrated and compared under multi-scales,to explore the evolution rules of calibrated parameters.