基于用户内容信息转移的社会网络链接预测研究
发布时间：2019-01-10 13:39
【摘要】：社会网络是由社会行为者即个人或者组织,社交关系以及行为者之间的其他社交交互组成的社会结构,在社会科学中用于研究个人、团体、组织甚至整个社会之间的关系。随着在线社交网络的发展,信息产生和传播的成本大大下降,信息的数量呈几何倍数的增长。这些网络中每天产生的巨大数据具有海量、高维、半结构化等明显的特征,这些特征可以直接反映人类社会的真实活动规律,所以社会网络逐渐成为多领域研究者的研究热点。社会网络分析最初作为社会学研究的一个分支,后来逐渐在数学、社会科学、人类学、生物学、通信科学等领域发展起来。社会网络具有高度动态性,可能导致节点和边在未来某个时刻出现或消失。因此,在社会网络链接预测中预测当前网络中缺失的边和新的或消失的未来网络中的边,对于挖掘社会网络中的未知信息和分析社会网络的演化是重要的。传统基于网络拓扑的节点相似性预测链接变化的方法很少考虑用户产生消息内容本身的特征,本文在传统社会网络链接预测方法的基础上引入用户产生的内容信息,利用转移熵量化用户对之间信息的转移作为用户之间相似性的一个特征,然后本文利用信息转移特征与拓扑特征的各种线性组合定义了三种链接预测方法。本文分别在基于LDA模型主题向量表示内容和分布式词向量表示内容下针对上述三种链接预测方法进行实验,验证信息转移对链接预测的影响并比较基于信息转移的链接预测方法与几个经典传统链接预测算法。在实验中发现了结合了信息转移的网络在社会网络链接预测中具有更加符合真实社交网络、更好的链接预测性能,在社会网络分析中具有一定的优势。
[Abstract]:Social network is a social structure composed of individual or organization, social relations and other social interactions between social actors. It is used in social sciences to study the relationships among individuals, groups, organizations and even the whole society. With the development of online social networks, the cost of information generation and dissemination has been greatly reduced, and the amount of information has increased exponentially. The huge data generated every day in these networks have the obvious characteristics of massive, high-dimensional, semi-structured and so on. These characteristics can directly reflect the real activities of human society, so the social network has gradually become the research hotspot of multi-field researchers. As a branch of sociological research, social network analysis has gradually developed in the fields of mathematics, social science, anthropology, biology, communication science and so on. Social networks are highly dynamic and may cause nodes and edges to appear or disappear at some point in the future. Therefore, it is important to predict the missing edges in the current network and the edges in the new or vanishing future networks in the prediction of social network links for mining unknown information in social networks and analyzing the evolution of social networks. The traditional method of node similarity prediction based on network topology rarely considers the characteristics of user-generated message content itself. This paper introduces user-generated content information based on the traditional social network link prediction method. The transfer entropy is used to quantify the information transfer between users as a feature of the similarity between users. Then three link prediction methods are defined by using various linear combinations of information transfer features and topological features. In this paper, the above three link prediction methods are tested based on LDA model subject vector representation and distributed word vector representation, respectively. The effect of information transfer on link prediction is verified and several classical link prediction algorithms based on information transfer are compared. In the experiment, it is found that the network combined with information transfer has more realistic social network and better link prediction performance in the social network link prediction, and it has some advantages in social network analysis.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09
本文编号：2406379
[Abstract]:Social network is a social structure composed of individual or organization, social relations and other social interactions between social actors. It is used in social sciences to study the relationships among individuals, groups, organizations and even the whole society. With the development of online social networks, the cost of information generation and dissemination has been greatly reduced, and the amount of information has increased exponentially. The huge data generated every day in these networks have the obvious characteristics of massive, high-dimensional, semi-structured and so on. These characteristics can directly reflect the real activities of human society, so the social network has gradually become the research hotspot of multi-field researchers. As a branch of sociological research, social network analysis has gradually developed in the fields of mathematics, social science, anthropology, biology, communication science and so on. Social networks are highly dynamic and may cause nodes and edges to appear or disappear at some point in the future. Therefore, it is important to predict the missing edges in the current network and the edges in the new or vanishing future networks in the prediction of social network links for mining unknown information in social networks and analyzing the evolution of social networks. The traditional method of node similarity prediction based on network topology rarely considers the characteristics of user-generated message content itself. This paper introduces user-generated content information based on the traditional social network link prediction method. The transfer entropy is used to quantify the information transfer between users as a feature of the similarity between users. Then three link prediction methods are defined by using various linear combinations of information transfer features and topological features. In this paper, the above three link prediction methods are tested based on LDA model subject vector representation and distributed word vector representation, respectively. The effect of information transfer on link prediction is verified and several classical link prediction algorithms based on information transfer are compared. In the experiment, it is found that the network combined with information transfer has more realistic social network and better link prediction performance in the social network link prediction, and it has some advantages in social network analysis.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09
【参考文献】
相关期刊论文 前1条
1 WANG Peng;XU BaoWen;WU YuRong;ZHOU XiaoYu;;Link prediction in social networks: the state-of-the-art[J];Science China(Information Sciences);2015年01期
相关硕士学位论文 前1条
1 魏超;社交网络中的链接预测研究[D];华中科技大学;2012年
，本文编号：2406379
本文链接：https://www.wllwen.com/guanlilunwen/ydhl/2406379.html
