Research On The Impacts Of Emergency Events On Human Communication Behavior Patterns

Researches on characteristics of human psychology and behavior under extreme events are of great importance in emergency management.Analysis of human spatiotemporal dynamics based on the datasets of large scale digital footprints is one of the research hotspots of the emerging computational social science,which provides a novel methodology for the quantitative study of the specified problem.Focusing on the non Poisson dynamics of human communication behavior,a number of empirical results and models that reveal the intrinsic mechanism of human dynamics have appeared in top academic journals.Some researchers are also concerned about the impact of emergencies on human communication behavior,and find some abnormal phenomena such as the sharp increase of call volume,the rise of movement predictability and the enhancement of social network connectivity,which provide the references for the further study of the evolution and intrinsic mechanism of the influence of emergency events on human behavior.In this thesis,we find out distinguish spatiotemporal statistical characteristics of emergency calling patterns that are different from conventional communication behavior based on the datasets of emergency calling and mobile phone calling records.Then we explore the intrinsic mechanism of emergency calling dynamics and propose the agent based models to validate the proposed mechanisms by simulation.Moreover,the urban regional security risk characteristics at the macro level emerged from the emergency calling behavior are defined to verify the macroscopic characteristics representation ability of the models.The main contributions are as follows:(1)Empirical analysis: We find that the emergency calling behavior is not completely random through the investigation of spatiotemporal characteristics of real dataset.In terms of the temporal patterns,we find that the distribution of inter-call durations decays as power law along with an exponential tail,and comparing with most of the normal communication activities.The activity of emergency calling shows more significant characteristics of burstiness and memory.In terms of the spatial patterns,we find that there is obvious randomness in the conversion of space sequence and the distributions of the displacement and the radius of gyration can be fitted by gaussian functions.However,the positions of emergency callings show a strong double center gathered characteristics relative to the home and work place at the aggregated level.There is certain memory effect in the conversion process although the proportion of the number of positions assigned to the two centers are almost equal.(2)Modeling: Based on the empirical study,we analyze the related mechanisms,and propose an event driven memory response model(EDMR)and a double center memory-effect random walk model(DCMRW)for the generation of emergency calling behavior with our algorithms.The simulation results indicate that our models can reproduce the patterns of real world emergency calling behavior well,which verify the mechanisms of the models.Then we analyze the influence of the parameters on the statistical properties of the simulation data by KL divergence.(3)Application: We propose a framework of multi spatial resolution regional networks based on emergency calling behavior,and define the security risk characteristics of urban areas at the macro level emerged from the emergency calling behavior such as the security risk index,regional hotspots,the origin and diversity of risks.After the work of model validation,we analyze the diversity of risk index and relation of areas in a metropolitan of China.Our work can provide reference for emergency management,such as optimizing the arrangement of emergency resources.