基于线性阈值模型的社交网络影响力最大化研究

发布时间：2019-01-25 19:00

【摘要】：近几年,社交网络在互联网中的地位越来越重要,已经被广泛地进行了研究,因为人们更愿意在社交网络中分享他们的想法和心情状态,社交网络中蕴藏着大量有价值的信息,利用社交网络可以进行许多的商业活动,例如广告、舆情分析、信息传播等。其中从社交网络中挖掘有限的一些用户,利用这些用户进行商品推广和信息传播正变得越来越热门,已经形成了一类研究问题—社交网络影响力最大化。社交网络影响力最大化是这样一类问题,在社交网络中识别一些最有影响力的人,这些人作为初始的传播信息的源头,可以将信息传播到最多的人。然而,现有的方法都忽略了社交网络中人的兴趣因素,这些方法和模型是不合理的。因为现实中人会有多个兴趣,并且对每个兴趣的敏感程度也不一样。另外,这些方法也忽略了要传播的信息的内容,因为不同背景的人对不同的信息表现也不一样,所以同样的人群对于不同的信息有着不同的影响力。本文针对已有的研究工作,指出了这些工作中的不足和缺陷,主要集中在已有的工作没有考虑到用户的兴趣因素,同时也没有考虑到要传播的信息的内容,以至于挖掘出来的有限的用户并不能够使传播信息的影响力最大化。本文解决了上述两个主要问题,结合之前的研究工作,对社交网络影响力最大化重新进行了定义,提出了携带兴趣组的社交网络影响力最大化的概念,设计了一种方法把社交网络里的兴趣组识别出来,并且结合兴趣组的概念,提出了一种新的衡量多兴趣组社交网络影响力的方法,最终提出了一个新奇的IING(Identifying Influential Nodes Greedy Algorithm)算法来计算最有影响力的用户,IING算法能够使挖掘到的一些用户作为初始的信息传播源时,信息能够被更多的人接受。最后,本文对提出的识别社交网络中的兴趣组的方法在真实的数据集上进行了实验,证明了方法的有效性。并且针对最终提出的IING算法进行了大量的实验验证,实验结果表明,本文提出的IING算法在时间上和效果上都优于现有的方法。
[Abstract]:In recent years, the status of social networks in the Internet has become increasingly important and has been extensively studied, because people are more willing to share their thoughts and mood states in social networks, which contain a lot of valuable information. Social networks can be used for many business activities, such as advertising, public opinion analysis, information dissemination, and so on. Among them, it is becoming more and more popular to mine a limited number of users from social networks, using these users to promote goods and spread information, which has formed a kind of research problem-maximizing the influence of social networks. Maximizing the influence of social networks is a problem in which some of the most influential people are identified, who, as the source of initial dissemination of information, can spread information to the largest number of people. However, the existing methods ignore the human interest in social networks, and these methods and models are unreasonable. Because in reality people have more than one interest, and the sensitivity to each interest is different. In addition, these methods also ignore the content of the information to be disseminated, because different people from different backgrounds have different information performance, so the same people have different influence on different information. Aiming at the existing research work, this paper points out the shortcomings and shortcomings of these work, which mainly focus on the fact that the existing work does not take into account the interest of the user, nor the content of the information to be disseminated at the same time. So that the limited users excavated can not maximize the impact of the dissemination of information. This paper solves the above two main problems, combines the previous research work, redefines the social network influence maximization, and puts forward the concept of the social network influence maximization with interest group. This paper designs a method to identify interest groups in social networks, and proposes a new method to measure the influence of multi-interest groups social networks by combining the concept of interest groups. Finally, a novel IING (Identifying Influential Nodes Greedy Algorithm) algorithm is proposed to calculate the most influential users. The IING algorithm can make the information accepted by more people when some users are used as the initial source of information propagation. Finally, the proposed method for identifying interest groups in social networks is tested on real data sets, and the effectiveness of the method is proved. The experimental results show that the proposed IING algorithm is superior to the existing methods in time and effect.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09

【参考文献】