【作者】 李博；

【导师】 杨静；

【作者基本信息】 哈尔滨工程大学 ， 计算机科学与技术， 2021， 硕士

【摘要】 近些年来,随着各式的社交平台的发展,以及在线社交的广泛普及,这类平台已经成为了大多数人生活中必不可少的一个组成部分,人们通过这样的途径去进行日常交流。社交网络也在不断地改变人们的生活习惯,人们更习惯于通过该方式去获取消息。在社交网络的研究领域中,非常关键的一环就是对网络中节点影响力的判断,因此,对影响力大小的分析是非常有必要的,该问题是节点影响力最大化的基础。其中K-shell算法是一种较为有效的影响力划分算法,该算法能够快速地在一个网络中将节点进行分层,并根据节点所在的层级判断节点的影响力大小,因此K-shell算法一直以来都是人们在社交网络领域中研究的热门。本文对于传统的K-shell算法对网络中节点的划分不确切的问题,提出了一种基于K-shell的社交网络节点影响力最大化算法,该算法基于K-shell的核心思想,利用在Kshell分解过程中的参数,例如迭代次数,节点层级等,并且综合考虑了网络中节点的位置属性以及邻域属性对节点影响力大小所产生的影响。以邻域节点在K-shell分解过程中的迭代轮次并结合当前节点到网络中心的距离来确定节点的位置属性,并且在计算节点对邻域的作用时,计算了邻居节点度和位置的共同作用,并最终通过位置属性和邻域属性两个方面来确定一个节点最终影响力,用本文提出的方法来验证改进算法排名的正确性。此外,种子节点集选取的过程中,充分考虑到了节点在传播时影响力的重叠问题,通过以节点的位置来对节点影响力进行不同程度的削减,解决了在影响力传播过程中的影响力重叠现象,有效消除了影响力浪费问题,最后通过结合以上两种方案提出一种影响力最大化算法。最后,在4组真实数据集上进行了相关的对比模拟实验,实验结果表明本文提出的改进算法相比于改进前以及相关其它方法有更高的准确性。并且通过实验可以看出,改进算法在各种规模的数据集中都有较为明显的传播效果提升。更多还原

【Abstract】 In recent years,with the development of various social platforms and the widespread popularity of online social networking,such platforms have become an indispensable part of most people’s lives,and people use such channels for daily communication.Social networks are also constantly changing people’s living habits,and people are more accustomed to obtaining news in this way.In the field of social network research,a very critical link is the judgment of the influence of nodes in the network.Therefore,it is necessary to analyze the magnitude of influence.This problem is the basis for maximizing the influence of nodes.Among them,the K-shell algorithm is a more effective influence division algorithm.The algorithm can quickly layer nodes in a network,and judge the influence of the nodes according to the level of the node.Therefore,the K-shell algorithm has been It has been a hot topic for people to study in the field of social networkingIn this paper,regarding the problem of inaccurate division of nodes in the network by the traditional K-shell algorithm,a K-shell-based algorithm for maximizing the influence of social network nodes is proposed.The parameters in the shell decomposition process,such as the number of iterations,node level,etc.,and comprehensively consider the influence of the location attribute of the node in the network and the neighborhood attribute on the influence of the node.The iterative rounds of the neighboring nodes in the K-shell decomposition process are combined with the distance from the current node to the network center to determine the position attributes of the nodes,and when calculating the effect of the nodes on the neighborhood,the degree and position of the neighboring nodes are calculated.Work together,and finally determine the final influence of a node through two aspects of location attribute and neighborhood attribute,and use the method proposed in this paper to verify the correctness of the improved algorithm ranking.In addition,in the process of selecting the seed node set,full consideration has been given to the overlap of the influence of nodes during propagation.The position of the node is used to reduce the influence of the node to varying degrees,which solves the influence in the process of influence propagation.The overlap phenomenon effectively eliminates the problem of wasting influence.Finally,an influence maximization algorithm is proposed by combining the above two schemes.Finally,a comparative simulation experiment is carried out on four real data sets.Experimental results show that the improved algorithm proposed in this paper is more accurate than the original algorithm and other related algorithms.Experiments show that the improved algorithm can improve the propagation effect in all kinds of data sets.更多还原