基于人类动力学的微博用户行为统计特征分析与建模研究

发布时间：2018-04-28 00:31

本文选题：微博 + 兴趣驱动　；参考：《北京邮电大学》2012年博士论文

【摘要】：随着移动通信技术和信息网络技术的发展与融合,互联网正式迈入Web3.0时代,也称移动互联网时代。该阶段以“应用服务”为中心,强调服务的移动性、泛在化、智能化、个性化与多元化。微博,以其实现了移动终端与互联网的无缝连接,以及易用、随意、即时、传播快等特点,在不到三年的时间内已发展成为重要的分享工具和自媒体传播平台,成为人们网上生活的重要组成部分。如何深入理解微博用户的行为,从而利用好微博平台,这是政府相关管理机构和网络营销企业共同关注的问题。在传统的人类行为研究中,人们常常假设人类行为在时间上是随机的,可以近似用泊松分布进行描述。近几年来,随着数据搜集和计算机处理能力的提高,越来越多关于人类动力学领域的实证研究表明,人类行为明显偏离了泊松过程。这些实证研究基本涵盖了传统互联网上的行为,但很少涉及移动互联网的研究。此外,行为内在驱动机制方面的研究成果较少,比较有影响力的主要有优先权选择、兴趣驱动、记忆影响等,但由于人类行为具有高度的复杂性,单一的驱动机制往往只能用于解释某些特定的人类动力学行为,不具备普适性。从模型的验证来看,已构建的理论模型能够生成特定范围的幂指数,但缺乏对参数的量化,给模型的实际应用带来制约。关于微博研究的国内外成果主要集中在微博信息传播机制,以新闻领域、传播领域的学者居多,其次是微博使用动机和微博营销的研究,微博用户行为的研究最少。研究方法多采用问卷调查、案例分析、理论解释等进行定性研究,少量关于用户行为的分析选取了微博网站的实际数据,但多停留于词频分析、相关性分析等基础统计。国外的研究多数采用定量分析和定性分析结合的方式,但在用户行为的研究深度上有待加强。针对目前研究中存在的空白和不足,本文选取国内影响最大的微博网站——新浪微博的用户作为研究对象,以人类动力学相关理论为基础,综合运用复杂网络、统计物理、概率论、应用统计学、管理学等多学科理论知识和方法,对用户的微博信息发布行为、转发行为和评论行为进行了深入的统计分析和建模研究。本文的主要研究结论和研究成果如下：第一,揭示用户微博信息发布行为的统计特征并构建理论模型。统计分析结果表明,用户的微博信息发布具有周期性、阵发性特征,无论是工作日还是周末,信息发布的高峰期都出现在上午12点左右,这不同于电子邮件、即时通信、手机的使用高峰期常常出现在上午10点左右。微博使用高峰期的滞后表明,微博信息基本与工作无关,只是利用碎片时间的一些短暂交流。微博的评论转发数与平均时间间隔在双对数坐标下存在较为显著的负相关关系。这表明,用户的微博信息发布行为受社会关注的影响。以统计分析结果为依据,本文提出了引入社会关注机制的兴趣驱动模型,并给出了特定情况下时间间隔分布的解析表达式,数值仿真结果与解析结果相一致。模型仿真表明,用户发布微博信息的时间间隔服从幂律分布,发布行为受兴趣驱动。兴趣越大,微博信息发布的时间间隔越短,而兴趣随时间呈周期性衰减趋势,社会关注能够有效减缓兴趣的衰减程度,激发用户在短期内高频率的发布微博信息。第二,揭示用户微博信息转发评论行为的统计特征并构建理论模型。统计分析结果表明,微博用户的转发数、评论数在双对数坐标下呈正相关关系,评论数的频数分布、转发数的频数分布以及转发评论数的频数分布均服从幂律分布,幂指数值在1和2之间。与评论行为相比,用户的转发行为更为频繁。此外,转发评论数与粉丝数关系密切,粉丝数越多的用户,其发布的微博信息更容易吸引大量的转发和评论。在此基础上,本文对BA模型以及节点批量到达的Poisson网络模型进行改进,引入微博影响力和优先连接机制,建立了用户的微博信息评论模型。对模型参数进行量化后的仿真结果与实际数据分析结果一致,模型能够较好的描述和解释用户的微博信息评论行为。仿真结果显示,微博影响力对评论数与微博信息连接的作用要高于优先连接。20%的明星微博信息所获的评论数占到了总评论数的近70%,彰显“意见领袖”的强大魅力。第三,揭示危机情境下用户微博行为的统计特征并构建理论模型。实证结果显示,用户发微博信息的时间间隔服从幂律分布,但与日常情境相比,用户发微博信息较为频繁,也更偏向使用PC客户端。其次,通过对微博信息内容进行分析,发现危机情境下博文长度并没有发生显著变化,但对于涉及危机情境主题的微博信息,大部分用户仅发布一次,而这类微博信息被转发和评论的概率也相对较高。此外,用户对于危机情境的关注在初期较高,此后随着事件的发展不断下降,且微博信息的发布数和转发评论数均随时间呈衰减趋势。根据危机情境下用户微博行为的统计分析结果,本文引入优先连接机制,并考虑微博数和转发评论数随事态衰减的因素,建立了危机情境下群体用户的微博信息转发评论模型。模型仿真结果表明,微博数的衰减程度以及优先连接的概率p的增大都会加剧转发评论数的聚集程度,使转发评论数频数分布的幂指数值变大,而转发评论数的衰减程度与幂指数变化呈现出先同向后反向的变化趋势。此外,基于用户微博行为的研究结论,本文分别从舆论引导和网络营销两个方面提出相关建议。舆论引导方面,本文强调监测时间、监测对象以及引导方式。微博营销方面,本文强调提高微博信息的影响力,同时激发粉丝用户发布相关微博信息的兴趣,并给出了企业自建微博帐号进行营销的具体实现路径。本文的创新点主要体现在以下四个方面： (1)从多个角度出发分析微博用户真实的行为数据,揭示出用户微博行为的统计特征。本文分别从周期性、阵发性、时间间隔分布、频数分布、相关性等多个方面进行统计分析,为建立相关的行为动力学模型提供依据； (2)建立了兴趣-社会关注共同驱动的用户微博信息发布模型。该模型除了描述兴趣衰减对用户微博发布时间间隔的作用外,更进一步指出社会关注是导致兴趣衰减程度发生变化的重要因素。社会认同和兴趣的共同作用能够有效激发用户高频率的发布微博信息； (3)建立了引入微博影响力和优先连接机制的用户微博信息评论模型。该模型量化了粉丝数对于微博信息获取更多评论的重要作用,刻画出明星微博信息和普通微博信息在转发评论行为中所扮演的不同角色,对转发评论热点的形成给出了合理解释； (4)建立了危机情境下群体用户的转发评论模型,揭示出外部环境因素对用户行为的影响。该模型量化了微博数衰减、转发评论数衰减等危机情境下用户微博行为的重要特征,并引入优先连接机制,对危机情境下的转发评论行为进行了描述。模型仿真结果与实证结果相符,能够较好的解释危机情境下用户行为随事态发展的变化趋势。
[Abstract]:With the development and integration of mobile communication technology and information network technology , the Internet has entered the era of Web3.0 , and also called mobile Internet era . The stage takes " application service " as the center , emphasizes the mobile terminal ' s seamless connection with the Internet , and becomes an important part of people ' s online life in less than three years .

In the traditional human behavior research , it is often assumed that human behavior is random in time and can be approximated with Poisson distribution . In recent years , with the improvement of data collection and computer processing capacity , more and more empirical studies on the field of human dynamics have shown that human behavior deviates significantly from Poisson process .

The results of micro - blog research mainly focus on micro - blog information transmission mechanism , in the field of news , the scholars in the field of communication are more and more , the second is the research of micro - blog usage motivation and micro - blog marketing . The research method uses questionnaire , case analysis , theoretical explanation and so on . The research method uses questionnaire , case analysis , theoretical explanation and so on .

Based on the theory of human dynamics , the author makes an in - depth statistical analysis and modeling research on the publishing behavior , forwarding behavior and comment behavior of user ' s microblog information .

The main findings and results of this paper are as follows :

First , the statistical characteristics of the publishing behavior of user ' s microblog information are revealed and the theoretical model is constructed . The statistical analysis results show that the user ' s microblog information publication has periodicity , paroxysmal characteristic , whether working day or weekend , the peak period of information distribution appears at about 12 a.m . , which is different from the e - mail and instant messaging .

Based on the statistical analysis results , this paper puts forward the interest driving model to introduce the social concern mechanism , and gives the analytic expression of time interval distribution under certain circumstances . The numerical simulation results are consistent with the analytic results . The model simulation shows that the time interval for the user to publish the microblog information is subject to the power law distribution , and the publishing behavior is driven by interest . The larger the interest is , the shorter the time interval of the microblog information release , and the social concern can effectively slow the attenuation degree of the interest and stimulate the user to publish the microblog information at a high frequency in a short time .

Secondly , the statistical characteristics of user ' s microblog information forwarding comment behavior are revealed and the theoretical model is constructed . The statistical analysis results show that the number of forwarding of micro - blog users , the frequency distribution of the number of comments and the frequency distribution of the number of forwarding comments follow the power law distribution , and the power exponent value is between 1 and 2 . Compared with the comment behavior , the user ' s forwarding behavior is more frequent . In addition , the number of forwarding comments is close to the number of fans , and the more users the number of fans is , the more users whose number is published can attract a large amount of forwarding and comment more easily .

On the basis of this , this paper improves the Poisson network model of the BA model and the node batch arrival , introduces the micro - blog influence and the priority connection mechanism , and establishes the user ' s microblog information comment model . The simulation result of the model parameter is consistent with the actual data analysis result . The simulation results show that the influence of the micro - blog influence on the connection of the comment number and the microblog information is higher than that of the priority connection .

Third , revealing the statistical characteristics of user ' s micro - blog behavior in crisis situation and constructing the theoretical model . The empirical results show that the time interval of user ' s micro - blog information is subject to power law distribution , but compared with the daily situation , the user ' s microblog information is more frequent and prefers to use PC client .

Based on the results of statistical analysis of user ' s micro - blog behavior under crisis situation , this paper introduces the priority connection mechanism , and considers the factors of micro - blog number and the number of forwarding comments along with the events . The model simulation results show that the degree of attenuation of micro - blog and the increase of probability p of priority connection can increase the aggregation degree of the number of forwarding comments .

In addition , based on the research conclusion of user ' s micro blog behavior , this paper puts forward relevant suggestions from the aspects of public opinion guidance and network marketing . In the aspect of public opinion guidance , this paper emphasizes the monitoring time , monitoring object and guidance method . In the aspect of micro blog marketing , this paper emphasizes the influence of micro blog information and the interest of arousing the user to publish relevant micro blog information , and gives the concrete realization path for the marketing of enterprise self - built micro blog account .

The innovation points of this paper are mainly embodied in the following four aspects :

( 1 ) analyzing the real behavior data of the microblog users from a plurality of angles , revealing the statistical characteristics of the user micro - blog behavior , and carrying out statistical analysis on a plurality of aspects , such as periodicity , paroxysmal , time interval distribution , frequency distribution , correlation and the like , to provide a basis for establishing the relevant behavioral dynamics model ;

( 2 ) Interest - social concern is established . This model further points out that social concern is an important factor leading to changes in the degree of interest attenuation , in addition to describing the role of interest attenuation on user micro - blog post - release interval , and the common role of social identity and interest can effectively stimulate the user ' s high - frequency publishing micro - blog information .

( 3 ) The user micro blog information comment model is established which introduces the micro blog influence and the priority connection mechanism . The model quantified the important function of the number of fans to get more comments on the microblog information , and the different roles played by the celebrity micro blog information and the general micro blog information in the forwarding comment behavior are carved out , and the formation of the forwarding comment hot spot is reasonably explained ;

( 4 ) The forwarding comment model of group users in crisis situation is established , and the influence of external environment factors on user behavior is revealed . The model has quantified the important characteristics of user ' s micro - blog behavior in crisis situations such as the attenuation of micro - blog , the attenuation of forwarding comments and the like , and introduces the priority connection mechanism to describe the forwarding comment behavior in crisis situation .

【学位授予单位】：北京邮电大学
【学位级别】：博士
【学位授予年份】：2012
【分类号】：F224;F49

【共引文献】