在线社交网络中代价限制的影响力传播最大化技术研究

发布时间：2018-05-07 14:45

本文选题：在线社交网络 + 信息传播　；参考：《国防科学技术大学》2014年硕士论文

【摘要】：随着QQ、微信等网络社交应用的迅速发展,在线社交网络中的信息传播技术和信息传播效果受到了越来越多的关注。信息传播的影响力最大化问题研究在社会网络中寻找具有给定节点数目的传播源节点集合,通过该集合中的节点能够使信息最终传播到网络中最广泛的人群。传统的影响力最大化问题采用覆盖人数来评价影响效果,并未考虑用户是否为信息的传播目标,评价结果并不准确。面向目标人群的信息传播要求首先对网络用户的特征进行分析,本文提出了基于PageRank的用户特征标签重要性分析技术UWTA。该技术利用用户之间的关系网络,在用户之间建立PageRank投票模型来分析计算不同的特征标签对于用户的重要性。特征标签的重要性大小是本文判断用户是否为传播目标以及最终计算信息传播效果的指标。在线社交网络中已有的信息传播模型在解决影响力最大化问题时具有很高的时间复杂度。为提高社交网络中信息传播模型的效率,本文对广泛使用的独立级联模型进行了分析研究,根据独立级联模型中信息传播的概率性特征提出了基于图精简的CGIC模型,提高了解决影响力最大化问题的效率。实际应用中,信息传播是有一定代价的,解决影响力最大化问题的传统方法并未考虑代价因素。本文设计了一种采用CGIC模型的贪心策略BTIDM。该策略以用户的代价预算为限制条件,以信息在目标用户上的传播质量作为评价指标,采用CGIC模型寻找对目标用户影响力最大的传播源节点集合。本文实验数据来自arXiv网站中的作者合作关系网并标注了作者领域标签。实验结果表明,本文提出的UWTA技术能够准确分析不同特征标签对用户的权重;在模型的时间复杂度方面,本文提出的CGIC具有更好的运行效率;在影响力传播的效果方面,本文提出的BTIDM方法能够在更好影响目标人群的同时把代价限制在更小范围。
[Abstract]:With the rapid development of online social applications, such as QQ and WeChat, more and more attention has been paid to the information dissemination technology and the effect of information dissemination in online social networks. The problem of maximizing the influence of Information dissemination; A set of nodes with a given number of nodes can be found in social networks, through which information can eventually be transmitted to the widest population in the network. The traditional influence maximization problem uses the number of people to evaluate the effect of impact, and does not consider whether the user is the target of information dissemination, so the evaluation result is not accurate. The information communication for the target population requires the analysis of the characteristics of the network users. This paper presents the importance analysis technology of user feature tags based on PageRank. Using the relationship network between users, this technique establishes PageRank voting model among users to analyze and calculate the importance of different feature tags for users. The importance of feature label is the index to judge whether the user is the target of propagation and to calculate the effect of information dissemination. The existing information dissemination models in online social networks have high time complexity in solving the problem of maximizing influence. In order to improve the efficiency of the information transmission model in social networks, the widely used independent cascade model is analyzed and studied in this paper. According to the probabilistic characteristics of information transmission in the independent cascade model, a graph-based simplified CGIC model is proposed. It improves the efficiency of solving the problem of maximization of influence. In practical application, information dissemination has a certain cost, and the traditional method to solve the problem of maximizing influence does not consider the cost factor. In this paper, a greedy strategy based on CGIC model is designed. The strategy takes the cost budget of the user as the limiting condition and the quality of the information spread on the target user as the evaluation index. The CGIC model is used to find the set of propagating source nodes that have the greatest influence on the target user. The experimental data in this paper come from the author's cooperation network on arXiv website and label the author's domain. The experimental results show that the proposed UWTA technology can accurately analyze the weight of different feature tags to users; in terms of the time complexity of the model, the CGIC proposed in this paper has better running efficiency; The BTIDM method proposed in this paper can not only better affect the target population, but also limit the cost to a smaller range.
【学位授予单位】：国防科学技术大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09

【参考文献】