基于社交网络的连接关系研究与应用

发布时间：2018-03-20 02:40

本文选题：社交网络　切入点：关系强度　出处：《北京邮电大学》2014年硕士论文　论文类型：学位论文

【摘要】：目前如Facebook、微博等在线社交网络服务(Online Social Network, OSN)已经成为互联网中的重要应用,用户数持续增长,成为很多人必不可缺的信息获取手段和社交途径。用户及其之间的连接关系是组成和维持社交网络的基本要素,本文以连接关系为研究对象,从连接强度、连接类型、连接动态性等角度对其进行深入研究。这一研究也可以为OSN领域的相关研究提供理论基础,如推荐系统、隐私保护等问题,帮助服务提供商提供更具有实际价值的服务。本文首先针对连接关系,调研了多个领域中的研究理论和成果,对其进行对比和分类,提出了基于OSN的用户连接关系研究的内容框架。随后,分别提出了基于多元逐步回归的连接强度测量算法(MSLR)和基于随机游走策略的连接类型识别算法(RW-RT),能够充分利用社交网络中的用户信息、交互信息和用户之间的依赖于好友关系形成的拓扑网络,准确的识别用户之间的关系强度和关系类型。利用新浪微博的中的真实用户数据,关系强度测量MSLR的准确度约为80%,连接类型算法RW-RT的测量准确度约为85%。证明了算法的有效性和准确性。对于用户连接关系的动态性问题,由于用户行为会导致用户关系随时间发生变化,因此本文从用户行为的时间模式为出发点,侧面反映连接关系的动态性。利用小波变换和动态时间弯曲的K-Medoids算法(WT-DKM),得到微博用户行为的典型时间模式。此外,本文还基于MSLR算法和RW-RT算法开发了一款新浪微博应用,能够自动测量用户之间的关系强度,并从好友分组的角度帮助用户自动管理自己的好友关系,证明了算法的实际价值。
[Abstract]:At present, online Social Network (OSNs), such as Facebook and Weibo, has become an important application in the Internet, and the number of users has continued to grow, and it has become an indispensable means of obtaining information and social networking for many people. The connection relationship between users and their connections is the basic element to form and maintain the social network. In this paper, the connection relationship is taken as the research object, from the connection strength, the connection type, This research can also provide a theoretical basis for the related research in the field of OSN, such as recommendation system, privacy protection and so on, and help service providers to provide more valuable services. In this paper, firstly, according to the connection relation, the research theories and achievements in many fields are investigated, compared and classified, and the content framework of the user connection relationship research based on OSN is put forward. The connection strength measurement algorithm based on multiple stepwise regression (MSLR) and the connection type recognition algorithm based on random walk strategy (RW-RTP) are proposed respectively, which can make full use of user information in social networks. Interactive information and the topological network between users that depend on the friend relationship, accurately identify the relationship between the user and the relationship between the intensity and type of relationship, using Sina Weibo in the real user data, The accuracy of relational strength measurement MSLR is about 80, the accuracy of connection type algorithm RW-RT is about 850.The validity and accuracy of the algorithm are proved. Since user behavior can cause user relationships to change over time, this paper starts with the time pattern of user behavior. By using wavelet transform and dynamic time bending K-Medoids algorithm WT-DKMN, the typical time pattern of Weibo's user behavior is obtained. In addition, this paper also developed a Sina Weibo application based on MSLR algorithm and RW-RT algorithm, which can automatically measure the relationship between users and help users manage their friends automatically from the point of view of friend grouping. The practical value of the algorithm is proved.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09

【参考文献】