计算机工程与应用 ›› 2021, Vol. 57 ›› Issue (10): 65-74.DOI: 10.3778/j.issn.1002-8331.2005-0148

• 理论与研发 • 上一篇    下一篇

基于信息扩散视角的虚拟社区用户影响力研究

卢开,周艳菊   

  1. 中南大学 商学院,长沙 410000
  • 出版日期:2021-05-15 发布日期:2021-05-10

Research on User Influence of Virtual Community Based on Perspective of Information Diffusion

LU Kai, ZHOU Yanju   

  1. Business College, Central South University, Changsha 410000, China
  • Online:2021-05-15 Published:2021-05-10

摘要:

节点排序研究领域中,少有研究考虑群聚效应下的群体规范对传播效率的影响,这可能导致用户影响力度量的准确性下降。针对这一问题,从信息扩散角度出发,借鉴创新扩散理论与Bass扩散模型,提出一种适用于虚拟社区网络的用户局部影响力度量模型CSA-LL(Cohesive Subgroup Analysis Based Local Leadership):基于凝聚子群挖掘与分析,定义子群内部信息扩散效率,并结合用户全局影响力,计算模型输出值作为节点排序的依据。爬取近期的豆瓣社区数据进行网络构建,使用AISAS模型等方法验证了该模型输出的用户比PageRank算法和Hits算法结果具有更强的营销能力。使用LT模型进一步验证了模型的有效性和子群信息扩散效率对用户传播能力存在正向影响。再使用多个虚拟社区网络数据集和IC模型,分别验证了模型鲁棒性与结论稳健性。

关键词: 创新扩散理论, 凝聚子群分析, 虚拟社区, 用户影响力, LT模型

Abstract:

In the field of node ranking research, few studies consider the influence of group norms on the propagation efficiency under the clustering effect, which may cause the accuracy of user influence measurement to decrease. In view of this problem, from the perspective of information diffusion, drawing on innovation diffusion theory and Bass diffusion model, a user local influence measurement model CSA-LL(Cohesive Subgroup Analysis Based Local Leadership)for virtual community networks is proposed:Based on the mining and analysis of condensed subgroups, the internal information diffusion efficiency of subgroups is defined, combined with the user’s global influence, the model output value is calculated as the basis for node ranking. Recent Douban community data are crawled for network construction, methods such as the AISAS model are used to verify that the users output by this model have stronger marketing capabilities than the results of the PageRank algorithm and Hits algorithm. The LT model is used to further verify the effectiveness of the model and the information diffusion efficiency of the subgroup has a positive effect on users’ communication ability. Multiple virtual community network data sets and IC models are used to verify the robustness of the conclusions.

Key words: innovation diffusion theory, condensed subgroup analysis, virtual communities, user influence, LT model