在线社会网络用户特征分析与建模研究
发布时间：2018-08-13 17:54
【摘要】：随着计算机技术与互联网技术的迅猛发展,各类在线社会网络不断涌现。在线社会网络具有参与自由、使用方便、信息传播速度快、互动性强等特点,吸引了大量用户的关注与参与,并已成为人们日常生活与工作中最为重要的信息交流平台之一。基于转发、评论等交互行为,用户之间可以建立起虚拟的社交关系,在一定程度上可以被认为是真实社会关系在网络世界的延伸。研究在线社会网络中存在的规律与现象对解决实际社会中相关问题具有重要的指导意义。用户作为在线社会网络的核心,其特征与在线社会网络拓扑结构的建立、用户类型划分、信息传播规律等研究内容密切相关。如何有效的利用在线社会网络用户特征进行分析研究对诸如舆情监控、精准化营销等多个应用领域具有重大的应用价值。本文以大规模真实在线社会网络数据为对象,通过研究与分析网络用户的各类特征对在线社会网络领域中几类重要研究问题展开了细致的工作。具体包括用户影响力评估、用户类型划分以及基于在线社会网络的人群健康研究。其中,影响力评估以及用户类型划分一直以来都是研究者关注的重点,研究成果具有使舆情监控更加高效,以及提高商业营销的精准性等诸多价值。基于在线社会网络的人群健康研究是近年来新兴的热点研究问题,网络平台从数据到方法上可以给相关研究提供一条崭新的问题解决途径,对传统方法能够起到有益的补充或者替代。本文主要研究内容与研究成果如下:第一,在分析用户连接拓扑结构、用户行为以及用户信息等特征的基础上,本文提出了一个基于用户多类特征的影响力评估方法。该方法针对如何合理使用用户属性度量用户影响力的问题,首先通过贝叶斯网络综合分析各类用户特征对用户影响力的影响,随后借鉴PageRank算法思想衡量邻接节点的连接关系会对用户影响力产生作用,从用户自身属性与邻接属性两个方面对用户影响力进行评估。该方法能够避免现有方法中单一指标评估对用户影响力反映不够全面以及采用各属性加权方法中属性或行为量纲不一致所造成的物理含义不清楚等问题。最后,基于新浪微博数据进行了相关实验,结果验证了本文方法的有效性。第二,在分析在线社会网络用户转发链的基础上,本文提出了一个网络用户区域交互模型。该模型以用户行为特征为基础,描述了用户与其他不同邻接距离用户之间的交互行为,能够真实体现在线社会网络用户之间的交互模式。随后,基于区域交互模型研究了在线社会网络用户类型划分方法,从用户行为与用户影响力范畴等角度更为真实的体现出用户所属类型及其在网络中所处的地位。划分方法通过区域交互模型计算用户对邻接节点以及非邻接节点的显性、隐性影响力,根据不同类型网络用户交互模式以及两类影响力的分布模式将网络用户划分为重要用户、普通用户以及异常用户。最后,实验结果表明本文提出的模型能够针对不同类型在线社会网络用户进行有效的识别,并且与现有方法相比能够更为有效的解决相关问题。第三,本文基于在线社会网络用户信息特征对现实社会中的人群健康状况展开了探索研究,在测量分析的基础上借鉴传染病模型思想提出了一个人群肥胖预测方法。在线社会网络用户信息中包含了大量与用户生活习惯、兴趣爱好以及用户情感、身体状况的内容。通过在线社会网络到真实社会的映射,现实中人群健康状况可以由这些用户发布的信息内容体现。在线社会网络海量用户信息数据及其开放性给人群健康相关研究提供了一个坚实的数据平台以及解决问题的新途径,与基于传统医疗机构数据研究方法相比具有更高的效率。本文以现有研究为基础提出了几类与肥胖相关的特征,并从在线社会网络用户中提取出相关信息,然后对每一个特征与区域人群肥胖比例的相关性进行了分析。随后利用与肥胖相关的特征作为肥胖人群变化的系数,对不同地区人群肥胖状况的发展趋势进行了预测。实验结果显示,文本筛选得到的几类特征与人群肥胖有着较为密切的关系,并且基于这些特征的预测方法也具有较为理想的有效性。第四,本文在已有研究基础上开发了一个基于用户特征的在线社会网络用户分析系统。该系统能够有效的识别微博网络中的重要用户、异常用户等,并且可以对不同地区用户的肥胖健康状况进行分析。系统功能与在线社会网络研究领域中诸如影响力评估、用户类型划分等研究问题密切相关,具有一定的实用价值。上述研究内容与成果体现出了在线社会网络用户特征的重要性,同时也体现出了本文的研究价值。
[Abstract]:With the rapid development of computer technology and Internet technology, various kinds of online social networks are emerging. Online social networks have the characteristics of free participation, convenient use, fast information dissemination and strong interaction, which attract a large number of users'attention and participation, and have become the most important level of information exchange in people's daily life and work. Based on the interactive behaviors of forwarding and commenting, users can establish virtual social relations, which can be regarded as the extension of real social relations in the network world to a certain extent. For the core of online social network, its characteristics are closely related to the establishment of online social network topology, the classification of users, the law of information dissemination and other research contents. In this paper, we focus on large-scale real-time online social network data, and do some detailed work on several important research issues in the field of online social network by studying and analyzing the characteristics of network users. The impact assessment and user type classification have always been the focus of researchers'attention. The research results have many values, such as making public opinion monitoring more efficient and improving the accuracy of commercial marketing. The main contents and achievements of this paper are as follows: Firstly, based on the analysis of the topological structure of user connections, user behavior and user information, a user multi-class feature is proposed in this paper. In order to solve the problem of how to use user attributes reasonably to measure user influence, this method firstly analyzes the influence of various user characteristics on user influence through Bayesian network, and then uses PageRank algorithm to measure the effect of the connection relationship between adjacent nodes on user influence. This method can avoid the problems that the single index evaluation does not reflect the user's influence comprehensively in the existing methods and the physical meanings caused by the inconsistency of attributes or behavior dimensions in each attribute weighting method are not clear. Finally, based on the Sina Weibo number Secondly, based on the analysis of the online social network user forwarding chain, this paper proposes a network user area interaction model, which describes the interaction between users and other users with different adjacent distances on the basis of user behavior characteristics. Subsequently, based on the regional interaction model, the user type partitioning method of online social network is studied. From the perspective of user behavior and user influence category, the user type and its status in the network are more truly reflected. The explicit and implicit influence of users on adjacent nodes and non-adjacent nodes are calculated. According to different types of network user interaction patterns and two types of influence distribution patterns, network users are divided into important users, ordinary users and abnormal users. Finally, the experimental results show that the proposed model can be used for different types of online communities. It can identify network users effectively and solve related problems more effectively than existing methods. Thirdly, based on the characteristics of online social network users'information, this paper explores and studies the health status of people in real society, and puts forward a population fertilizer by referring to the idea of infectious disease model on the basis of measurement and analysis. Fat prediction method. User information on online social networks contains a lot of information about users'habits, interests, emotions and physical conditions. By mapping online social networks to the real world, people's health status can be reflected in the information released by these users. Information data and its openness provide a solid data platform and a new way to solve the problem for population health-related research. It is more efficient than traditional medical institution data research methods. Based on the existing research, this paper proposes several obesity-related characteristics and extracts them from online social network users. The correlation between each feature and the obesity ratio of the population in different regions was analyzed. Then, the trend of obesity in different regions was predicted by using the obesity-related characteristics as the coefficients of obesity change. Fourthly, an online social network user analysis system based on user characteristics is developed on the basis of existing research. The system can effectively identify important users, abnormal users and so on in the microblog network. The system function is closely related to the research issues in the field of online social network research, such as impact assessment, user type classification and so on, and has certain practical value. The research value of this article is given.
【学位授予单位】：西安理工大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP393.09
本文编号：2181744
[Abstract]:With the rapid development of computer technology and Internet technology, various kinds of online social networks are emerging. Online social networks have the characteristics of free participation, convenient use, fast information dissemination and strong interaction, which attract a large number of users'attention and participation, and have become the most important level of information exchange in people's daily life and work. Based on the interactive behaviors of forwarding and commenting, users can establish virtual social relations, which can be regarded as the extension of real social relations in the network world to a certain extent. For the core of online social network, its characteristics are closely related to the establishment of online social network topology, the classification of users, the law of information dissemination and other research contents. In this paper, we focus on large-scale real-time online social network data, and do some detailed work on several important research issues in the field of online social network by studying and analyzing the characteristics of network users. The impact assessment and user type classification have always been the focus of researchers'attention. The research results have many values, such as making public opinion monitoring more efficient and improving the accuracy of commercial marketing. The main contents and achievements of this paper are as follows: Firstly, based on the analysis of the topological structure of user connections, user behavior and user information, a user multi-class feature is proposed in this paper. In order to solve the problem of how to use user attributes reasonably to measure user influence, this method firstly analyzes the influence of various user characteristics on user influence through Bayesian network, and then uses PageRank algorithm to measure the effect of the connection relationship between adjacent nodes on user influence. This method can avoid the problems that the single index evaluation does not reflect the user's influence comprehensively in the existing methods and the physical meanings caused by the inconsistency of attributes or behavior dimensions in each attribute weighting method are not clear. Finally, based on the Sina Weibo number Secondly, based on the analysis of the online social network user forwarding chain, this paper proposes a network user area interaction model, which describes the interaction between users and other users with different adjacent distances on the basis of user behavior characteristics. Subsequently, based on the regional interaction model, the user type partitioning method of online social network is studied. From the perspective of user behavior and user influence category, the user type and its status in the network are more truly reflected. The explicit and implicit influence of users on adjacent nodes and non-adjacent nodes are calculated. According to different types of network user interaction patterns and two types of influence distribution patterns, network users are divided into important users, ordinary users and abnormal users. Finally, the experimental results show that the proposed model can be used for different types of online communities. It can identify network users effectively and solve related problems more effectively than existing methods. Thirdly, based on the characteristics of online social network users'information, this paper explores and studies the health status of people in real society, and puts forward a population fertilizer by referring to the idea of infectious disease model on the basis of measurement and analysis. Fat prediction method. User information on online social networks contains a lot of information about users'habits, interests, emotions and physical conditions. By mapping online social networks to the real world, people's health status can be reflected in the information released by these users. Information data and its openness provide a solid data platform and a new way to solve the problem for population health-related research. It is more efficient than traditional medical institution data research methods. Based on the existing research, this paper proposes several obesity-related characteristics and extracts them from online social network users. The correlation between each feature and the obesity ratio of the population in different regions was analyzed. Then, the trend of obesity in different regions was predicted by using the obesity-related characteristics as the coefficients of obesity change. Fourthly, an online social network user analysis system based on user characteristics is developed on the basis of existing research. The system can effectively identify important users, abnormal users and so on in the microblog network. The system function is closely related to the research issues in the field of online social network research, such as impact assessment, user type classification and so on, and has certain practical value. The research value of this article is given.
【学位授予单位】：西安理工大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP393.09
【参考文献】
相关期刊论文 前1条
1 毛佳昕;刘奕群;张敏;马少平;;基于用户行为的微博用户社会影响力分析[J];计算机学报;2014年04期
，本文编号：2181744
本文链接：https://www.wllwen.com/guanlilunwen/ydhl/2181744.html
