图书情报工作 ›› 2021, Vol. 65 ›› Issue (19): 142-150.DOI: 10.13266/j.issn.0252-3116.2021.19.014

• 综述述评 • 上一篇    

多节点多关系的混合网络社团划分研究综述

蒋璐1,2, 陈云伟1,2   

  1. 1 中国科学院成都文献情报中心科学计量与科技评价研究中心(SERC) 成都 610041;
    2 中国科学院大学经济与管理学院图书情报与档案管理系 北京 100190
  • 收稿日期:2021-04-06 修回日期:2021-08-02 出版日期:2021-10-05 发布日期:2021-10-09
  • 通讯作者: 陈云伟(ORCID:0000-0002-6597-7416)),研究员,硕士生导师,通讯作者,E-mail:chenyw@clas.ac.cn
  • 作者简介:蒋璐(ORCID:0000-0002-5514-9860),硕士研究生。
  • 基金资助:
    本文系国家社会科学基金项目"用于科学结构分析的混合网络的社团研究"(项目编号:19XTQ012)研究成果之一。

A Review of Community Detection in Hybrid Networks with Multiple Nodes and Multiple Relationships

Jiang Lu1,2, Chen Yunwei1,2   

  1. 1 Scientometrics&Evaluation Research Center(SERC), Chengdu Library and Information Center of Chinese Academy of Sciences, Chengdu 610041;
    2 Department of Library Information and Archives Management, School of Economics and Management, University of Chinese Academy of Sciences, Beijing 100190
  • Received:2021-04-06 Revised:2021-08-02 Online:2021-10-05 Published:2021-10-09

摘要: [目的/意义] 旨在对多节点多关系混合网络社团划分方法进行梳理,探析现有社团划分方法存在的问题及面临的困难,预见未来的发展趋势。[方法/过程] 对近年来有关多节点类型、多关系类型混合网络的社团划分方法研究工作进行系统梳理,从基于概率生成模型、元路径、种子节点、扩展模块度以及混合网络同构5个方面对划分方法进行阐述,归纳混合网络社团划分常用的评估指标:标准化互信息NMI、调整兰德指数ARI和模块度Q,指出社交媒体、学术网络、欺诈检测3个应用场景。[结果/结论] 归纳出多节点多关系混合网络社团划分方法的适用性及优缺点,揭示了当前发展面临的挑战,为后续的混合网络分析研究提供新的视角,并展望今后可能进一步拓展的相关研究方向。

关键词: 多节点, 多关系, 混合网络, 社团, 评估

Abstract: [Purpose/significance] By sorting out the community detection methods of multi-node and multi-relationship hybrid network, we can analyze the problems and difficulties existing in the community detection methods, and predict the development trend in the future.[Method/process] In this paper, the methods of multi-node type and multi-relation type hybrid network community detection were systematically reviewed, and were described from five aspects:based on probabilistic generation model, meta-path, seed node, expansion modularity and isomorphism of hybrid networks. This paper summarized the commonly used evaluation indicators for community detection in hybrid networks:Standardized Mutual Information(NMI), Adjusted Rand Index(ARI) and Modularity Q, and pointed out three application scenarios of social media, academic network and fraud detection.[Result/conclusion] This paper summarizes the applicability, advantages and disadvantages of the community detection methods of multi-node and multi-relationship hybrid network, reveals the challenges faced by the current development, provides a new perspective for the subsequent hybrid network analysis and research, and looks forward to the related research directions that may be further expanded in the future.

Key words: multi-node and multi-relation, hybrid network, community, evaluation

中图分类号: