社交团体的网络影响力最大化研究

发布时间：2018-01-07 01:34

本文关键词：社交团体的网络影响力最大化研究　出处：《华中科技大学》2015年硕士论文　论文类型：学位论文

【摘要】：病毒式营销是社交网络中重要的应用,现实场景中,社交网络中的用户通常属于某个具有特定组织结构的社交团体,因此如何选择给定数量的团体,基于所有团体内节点之间的相互信息传播,使得网络接收信息的用户数量最多,对于充分利用社交网络传播特性进行产品营销具有重要意义。对于社交团体的网络影响力最大化问题,在团体层面模拟网络中信息传播具有更好的拓展性,设计了模拟团体信息传播过程的GLISM模型,GLISM模型包括:基于团体的网络图、描述网络图上信息传播的信息传播规则。同时,给出了GLISM模型中团体集合信息传播范围的计算方法,并讨论了GLISM中信息传播范围函数的性质。设计基于GLISM模型的贪心算法GLISMGreedy,GLISMGreedy算法利用GLISM模型计算团体的信息传播范围边际效益,并结合GLISM模型的次模性质使得效率得到提升。为得到拓展性更好的算法,基于GLISM模型构建的网络图,设计了启发式的IR-DU算法,IR-DU算法以团体影响力作为团体信息传播范围边际效益的估值,并得到团体影响力初始值的计算方法,算法使用团体影响力作为选择种子团体的标准,并在选择种子团体过程中对团体影响力重新计算以减少团体影响力的覆盖重合。最后通过实验分析算法的效果,实验选择NetHEPT和DBLP数据集,以信息传播范围和时间效率作为评估标准,选取基本贪心算法、Topcgo算法以及基于度的启发式算法进行对比,对实验结果进行分析,实验结果表明GLISMGreedy算法和IR-DU算法得到的信息传播范围非常接近基本贪心算法,但时间效率能够有效提升。GLISMGreedy算法得到的信息传播范围要大于IR-DU算法,但IR-DU算法的效率更高。同时与Topcgo算法相比,GLISMGreedy算法和IR-DU算法能得到更大的信息传播范围,稳定性更好,且IRDU算法的时间效率要优于Topcgo算法。
[Abstract]:Viral marketing is an important application in the social network, real scene, social network users usually belong to a specific organizational structure of social groups, so how to select a given number of groups, mutual information communication between all nodes within the group based on the network information receiving the largest number of users, is of great significance for the make full use of the propagation characteristics of social network product marketing. The network social group influence maximization problem, in the team level simulation information dissemination network has better expansibility, the design of GLISM model group information dissemination process of the GLISM model include: network diagram group based on the description of information dissemination rules of information dissemination network map at the same time, gives the calculation method of GLISM model group collection of information spread, and discuss the scope of dissemination of information in GLISM The nature of the function. The design of the greedy algorithm GLISMGreedy based on the GLISM model, the marginal benefit of the GLISMGreedy algorithm using GLISM model to calculate the group's scope of dissemination of information, combined with the modular nature of the GLISM model makes the efficiency has been improved. In order to get better development of algorithm, network construction of GLISM model based on the heuristic IR-DU algorithm design, IR-DU the algorithm group influence as the marginal benefit of community information dissemination scope and calculation method of valuation, group influence the initial value of the algorithm, using the group influence as the seed selection group standards, and in the process of selecting seed group to group influence re calculated to cover the coincidence reduce influence. The experimental group analysis of the effect of algorithm, experiment select the NetHEPT and DBLP data sets, the information dissemination scope and time efficiency as evaluation criteria, selection of basic corruption Mental arithmetic method, Topcgo algorithm and heuristic algorithm based on the degree of comparison, the analysis of experimental results, the experimental results show that GLISMGreedy algorithm and IR-DU algorithm of the information transmission range is very close to the basic greedy algorithm, but the efficiency of.GLISMGreedy algorithm can effectively enhance the dissemination of information is greater than the scope of the IR-DU algorithm, but the efficiency of IR-DU algorithm is higher at the same time. Compared with the Topcgo algorithm, GLISMGreedy algorithm and IR-DU algorithm can obtain more information spread, better stability, and the IRDU time efficiency of the algorithm is superior to Topcgo algorithm.

【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：G206

【相似文献】