突发事件对人类通信行为模式影响的研究

发布时间：2018-06-10 20:35

本文选题：计算社会学 + 人类行为动力学　；参考：《天津大学》2016年博士论文

【摘要】：突发事件下人类应急心理与行为特性是应急管理基础研究关注的重要科学问题,而基于电子足迹数据分析人类行为时空动态特性是新兴的计算社会科学的研究热点之一,为该问题的定量研究提供了新的方法。关注于人类行为的非泊松动力学特性,近几年国内外一些研究机构的学者们在国际高水平学术期刊上发表了一系列关于人类通信行为时空统计特性的实证分析结果,并构建了相关的动力学模型解释其内在的驱动机制。同时,部分研究人员也开始关注突发事件对人类行为的影响,发现了通信量激增、移动可预测性升高、社交网络连通性增强等现象,为进一步研究突发事件影响下的人类行为模式及其动力学特性提供了参考。本文利用人类报警记录及常规通信数据集,发现了报警行为有别于常规通信行为的时空统计特征,分析了报警行为的动力学生成机制并构建了基于Agent的生成模型进行仿真验证,继而定义报警行为涌现出的宏观层面的城市区域安全风险特征,验证了模型的宏观特征再现能力。主要工作及创新点如下:(1)统计分析:通过对报警行为时空统计特性的分析发现报警行为并非完全随机,在时间模式上,发现报警行为序列的间隔时间概率分布呈现出幂律加近似指数尾的统计特性,其阵发性及记忆性相比大部分常规通信行为要高;在空间模式上,发现空间序列的转换具有一定的随机性,其单步位移和回转半径的分布可以由高斯分布拟合,但在群体层面下报警位置相对于家和工作地点呈现出较强的双中心聚集特征,虽然在两个中心的占比均衡,但转换过程中具有一定的记忆效应。(2)模型构建:基于报警行为统计特性的研究结果,对比分析了相关的动力学生成机制,在时间和空间两个侧面分别提出了事件驱动记忆响应(EDMR,Event Driven Memory Response)模型和双中心记忆性随机游走(DCMRW,Double Center Memory-effect Random Walk)模型,给出了报警行为时空序列的生成算法,仿真实验表明模型可以较好的重现报警行为的时空统计特性,验证了生成机制的有效性,并基于KL散度分析了模型中相关参数对仿真数据统计特性的影响。(3)领域应用:构建基于报警行为的多空间分辨率城市区域关联网络,定义了基于报警行为涌现出的宏观层面的城市区域安全风险特征,包括城市区域安全风险指数、热点区域、风险来源及其多样性等,在验证生成模型再现宏观特征有效性的基础上分析了中国某大都市区域安全风险态势及其多样性、关联性等特征,为优化资源部署等应急管理领域的应用提供参考。
[Abstract]:The characteristics of human emergency psychology and behavior is an important scientific issue in the basic research of emergency management, and it is one of the new research hotspots in computational social science to analyze the space-time dynamic characteristics of human behavior based on the electronic footprint data. It provides a new method for quantitative study of this problem. In recent years, scholars from some research institutions at home and abroad have published a series of empirical results on the space-time statistical characteristics of human communication behavior in international high-level academic journals. The related dynamic model is constructed to explain its internal driving mechanism. At the same time, some researchers have begun to pay attention to the impact of unexpected events on human behavior, and found that traffic surge, mobile predictability increased, social network connectivity increased, and so on. It provides a reference for the further study of the human behavior model and its dynamic characteristics under the influence of unexpected events. In this paper, using human alarm records and conventional communication data sets, the spatio-temporal statistical characteristics of alarm behavior different from conventional communication behavior are found, and the dynamic generation mechanism of alarm behavior is analyzed, and the generating model based on Agent is constructed for simulation verification. Then, the macro-level risk characteristics of urban regional security emerged from alarm behavior are defined, which verifies the model's ability to reproduce macro-features. The main work and innovations are as follows: statistical analysis: through the analysis of the space-time statistical characteristics of alarm behavior, it is found that alarm behavior is not completely random, and in time mode, It is found that the interval probability distribution of alarm behavior sequence shows the statistical characteristics of power law plus approximate exponential tail, and its paroxysmal and memory behavior is higher than that of most conventional communication behaviors. It is found that the spatial sequence transformation is random, and the distribution of the single step displacement and the radius of rotation can be fitted by the Gao Si distribution, but the alarm position at the group level shows a strong bicentric clustering characteristic relative to the home and the work place, and the distribution of the single step displacement and the radius of rotation can be fitted by the distribution of the single step displacement and the radius of rotation. Although the ratio of the two centers is balanced, there is a memory effect in the conversion process. Based on the research results of the statistical characteristics of alarm behavior, the dynamic generation mechanism is compared and analyzed. In this paper, the event driven memory response (EDMRN) model and the dual center memory random walk model (DCMRW / double Center memory effect Walk) are proposed in both temporal and spatial aspects, respectively, and the algorithm for generating temporal and spatial sequences of alarm behavior is presented. The simulation results show that the model can reproduce the statistical characteristics of alarm behavior in time and space, and the validity of the generation mechanism is verified. Based on the KL divergence, the paper analyzes the influence of the related parameters on the statistical characteristics of the simulation data. The application of the model is as follows: the multi-spatial resolution urban area network based on alarm behavior is constructed. This paper defines the characteristics of urban regional security risk based on the emergence of alarm behavior, including urban regional security risk index, hot spots, risk sources and their diversity, etc. Based on the validation of the validity of the generated model to reproduce the macro features, this paper analyzes the characteristics of the regional security risk situation and its diversity and relevance in a metropolis in China, which provides a reference for the application of emergency management such as optimizing the deployment of resources.
【学位授予单位】：天津大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：D63;TP391.9

【相似文献】