Computer Engineering and Applications ›› 2021, Vol. 57 ›› Issue (23): 98-105.DOI: 10.3778/j.issn.1002-8331.2011-0058

• Big Data and Cloud Computing • Previous Articles     Next Articles

Research on Heterogeneity of Attention Flow Network Based on Structural Entropy

MA Manfu, GUO Chenbiao, LI Yong, ZHANG Zhongying, ZHANG Qiang, WANG Changqing   

  1. 1.College of Computer Science and Engineering, Northwest Normal University, Lanzhou 730070, China
    2.DNSLAB, China Internet Network Information Center, Beijing 100190, China
  • Online:2021-12-01 Published:2021-12-02

基于结构熵的注意力流网络异构性研究

马满福,郭晨彪,李勇,张钟颖,张强,王常青   

  1. 1.西北师范大学 计算机科学与工程学院,兰州 730070
    2.中国互联网络信息中心 互联网基础技术开放实验室,北京 100190

Abstract:

Structural entropy, as an important measure of the disorder degree of complex networks, reflects the heterogeneity of the structure within the network. Traditional structural entropy only pays attention to the “node” and “edge” in the network structure when describing the heterogeneity of complex networks, and there are deficiencies in characterizing the heterogeneity of the attention flow network structure. In this regard, this paper constructs an attention flow network based on online click behavior data. On the basis of the traditional network structure entropy, with the website’s edge weight, the total stay time of the website and other network characteristics comprehensively considered, the structure entropy model is defined. Furthermore, the comprehensive power of the website is calculated from the website’s flow intensity, the ability to attract attention and other indicators, and an attention flow network heterogeneity measurement algorithm ANSE is proposed. Experimental results show that the structural entropy proposed in this paper can effectively reflect the structural characteristics of the attention flow network, accurately measure the differences between websites in the attention flow network, analyze the importance of the websites.?The influence ranking proves the superiority and effectiveness of this algorithm by comparing it with traditional classic algorithms.

Key words: complex network, attention flow network, structural entropy, network heterogeneity, website importance

摘要:

结构熵作为复杂网络无序程度度量的重要手段,反映了网络内结构的异质性。传统结构熵在刻画复杂网络异构性时只关注网络结构中的“点”和“边”,表征注意力流网络结构的异构性特征时存在不足。对此,基于在线点击行为数据构建注意力流网络,在传统网络结构熵的基础上,综合考虑站点的边权重、站点的总停留时长等网络特征属性,定义了结构熵模型。进而,从站点的流强度、吸引注意力的能力等指标计算站点综合力,提出了注意力流网络异构性度量算法ANSE。实验结果表明,提出的结构熵可以有效地反映注意力流网络的结构特征,准确地度量注意力流网络中站点之间的差异性,分析站点重要性排序,通过和传统经典算法对比,在站点影响力排名上证明了该算法的优越性和有效性。

关键词: 复杂网络, 注意力流网络, 结构熵, 网络异构性, 站点重要性