Research On Temporal-spatial Patterns Of Human Mobility On Large Geographical Scales

The study on the temporal and spatial patterns of human behavior is the present focus of complex system research.Deeply understanding the spatiotemporal complexity of human behavior has contributed to explaining the internal dynamics of some complex phenomena.Still,it can provide new perspectives and approaches,and provide crucial support in behavioral prediction,information recommendation,transportation planning,and epidemic control,etc.Therefore,the empirical study on human behavior can bring great application potential.In recent years,the spatiotemporal characteristics of human behavior have been widely studied.Some models have been proposed to explain the hidden mechanism,but it is still far from fully understanding and explaining the properties of human behavior.This study focusses on spatiotemporal characteristics of human behavior on both collective and individual levels on large geographical scales.Empirical data is collected when mobile phone(terminal)users browse websites.It records the information of websites browse,mobile towers visit and the associated time data.Firstly,the data of individuals moving on large scales is filtered.Corresponding temporal and spatial characteristics are subsequently studied and compared with those of previous studies.Meanwhile,since the data set contains the information of both websites browse and mobile towers visit,this study analyzes the behavioral phenomenon of websites browse.Among the statistical results,many non-Markovian scaling features have been obtained,e.g.,the distribution of interval time follows a power law,the distribution of the numbers of tower visit is power-law.Moreover,a detailed analysis is given for anomalies that do not satisfy the standard power law or exponential distribution.Subsequently,based on the associated rule of the users,if two participants visit the same tower within one hour,then the related two individuals are defined connected.Thus,a undirected weighted network of the entire data set is established,and visualized.Based on the study of the network characteristics,it is found that the network is scale-free,in which the degree distribution of nodes,the weight distribution of nodes and the weight distribution of edges both follow a power law.Still,the cluster coefficient converges to a constant 0.55.Finally,in terms of the dataset of human mobility on large geographical scales,this study uses two existing spatial models of human behavior,i.e.,continuous-time random walk and individual behavioral(priority return and exploration)models to investigate the distributions of dwelling time,times of towers visit,entropy,and predictability.The results show that these two models neither can explain the distributions of entropy and predictability.Furthermore,this study proposes a novel human mobility model on large scales,which can better reproduce the distributions of information entropy and predictability.Thus,the effectiveness is verified.This study also exhibits the comparisons between empirical results and numerical studies under distinct parameter cases.Moreover,the results are obtained theoretically with analytical deviations,including the power-law feature of activity times,the power-law distribution of towers visit,the power-law relationship between the frequency of towers visit and their rank,etc.