电子学报 ›› 2020, Vol. 48 ›› Issue (3): 582-589.DOI: 10.3969/j.issn.0372-2112.2020.03.023

• 学术论文 • 上一篇    下一篇

社区环境下基于节点交互和主题的影响力计算模型

王大刚1, 钟锦1,2, 吴昊1   

  1. 1. 合肥师范学院计算机学院, 安徽合肥 230601;
    2. 中国科学技术大学计算机学院, 安徽合肥 230600
  • 收稿日期:2019-03-06 修回日期:2019-07-03 出版日期:2020-03-25
    • 作者简介:
    • 王大刚 男,1982年10月出生,安徽肥东人,讲师,2008年毕业于安徽大学获硕士学位,主要研究方向:数据挖掘,智能信息处理,Web与数据库E-mail:wdgx621@163.com;钟锦 男,1973年6月出生,安徽舒城人,教授,分别于2004和2008在合肥工业大学和中国科学技术大学获硕士和博士学位,主要研究方向:人工智能,博弈论,智能信息处理;吴昊 男,1983年5月出生,安徽舒城人,副教授,研究方向:智能视频处理与分析
    • 基金资助:
    • 安徽省自然科学基金 (No.1708085QF157); 安徽省高校优秀青年人才支持计划 (No.gxyq2017050); 安徽省教育教学委托研究项目 (No.2018jyxm1470); 国家大学生创新创业项目 (No.201914098034)

Influence Calculation Model Based on Node Interaction and Topic in Community Environment

WANG Da-gang1, ZHONG Jin1,2, WU Hao1   

  1. 1. School of Computer Science and Technology, Hefei Normal University, Hefei, Anhui 230601, China;
    2. School of Computer Science and Technology, University of Science and Technology of China Hefei, Anhui 230600, China
  • Received:2019-03-06 Revised:2019-07-03 Online:2020-03-25 Published:2020-03-25
    • Supported by:
    • Natural Science Foundation of Anhui Province (No.1708085QF157); Excellent Young Talents Project in Higher Education of Anhui Province (No.gxyq2017050); Research Project Commissioned by Department of Education of Anhui Province (No.2018jyxm1470); National College Students Innovation and Entrepreneurship Training Program (No.201914098034)

摘要: 为解决现有算法对社交网络节点影响力计算准确度不高的问题,本文整合节点不同维度信息,综合考虑节点在多个主题社区上的主题分布向量,提出一种新的节点影响力计算模型.模型首先将主题相关性作为先验信息;然后利用混合隶属度随机块(Mixed Membership Stochastic Block)模型表达节点间的交互关系,用主题模型学习主题内容;最后结合全局拓扑关系迭代计算节点的全局影响力.本文选取社交网络数据,以P@N、MAP等作为评价指标同现有主流算法进行比较.实验结果显示,本文算法有效提升了影响力节点识别的准确度和排名的有效性.

关键词: 主题, 影响力, 混合隶属度随机块, 先验

Abstract: To solve the accuracy problems of the existing algorithms in calculating the influence of social network nodes, by integrating different dimension information of nodes, and considering the topic distribution vector of nodes on multiple communities, a new model is proposed. It first regards the correlation between topics as the prior information, then uses the mixed membership stochastic block (MMSB) model to express the interaction among nodes, learns topic contents using topic model, and finally, iteratively calculates the global influence of nodes with global topological relationship. We select data from social networks, use P@N, MAP, etc., as the evaluation indicators, and compare the proposed algorithm with the existing mainstream algorithms. The experimental results show that our algorithm significantly improves the identification accuracy of influential nodes and the validity of ranking.

Key words: topic, influence, mixed membership stochastic block, prior

中图分类号: