面向多维属性融合的加权网络结构洞节点发现算法  

Weighted Network Structural Hole Node Discovery Algorithm for Multi-Dimensional Attribute Fusion

作　　者：王文涛 刘彦飞 毛博文 余成波 WANG Wentao;LIU Yanfei;MAO Bowen;YU Chengbo(School of Electrical and Electronic Engineering,Chongqing University of Technology,Chongqing 400054,China;College of Intelligence and Computing,Tianjin University,Tianjin 300072,China;Department of Information Security,Chongqing Police College,Chongqing 401331,China)

机构地区：[1]重庆理工大学电气与电子工程学院,重庆400054 [2]天津大学智能与计算学部,天津300072 [3]重庆警察学院信息安全系,重庆401331

出　　处：《信息网络安全》2024年第8期1265-1276,共12页Netinfo Security

基　　金：重庆市自然科学基金创新发展联合基金[CSTB2023NSCQ-LMX0014];重庆市教育委员会科学技术研究项目[KJZD-K202201701]。

摘　　要：在大规模复杂网络空间中,快速识别结构洞节点对于病毒和舆情的传播控制具有重要意义。针对现有识别结构洞节点的方法在网络结构发生变化时,识别精度不高的问题,文章基于多维属性映射与融合,提出一种结合邻接信息熵与邻接中心性的结构洞节点识别算法。该算法将加权的邻接信息熵作为邻居节点的信息量,使用邻接中心性度量节点传播这些邻居节点信息量的重要性,通过将结构洞节点的局部属性表示为节点传播信息的能力,识别网络中的关键结构洞节点。实验结果表明,在不同网络规模和网络结构的数据集下,该算法的ξ、τ和网络平均信息熵3个评估指标的总得分分别为0.470、1.679和4.027,优于现有算法,具有更优越和稳定的性能,且将该算法应用于大规模网络中仍然具有较低的时间成本。In large-scale complex network spaces,quickly identifying structural hole nodes is of great significance for controlling the spread of viruses and public opinion.Aiming at the problem that the existing methods for identifying structural hole nodes have low recognition accuracy when the network structure changes,this paper proposed a structural hole node recognition algorithm.The algorithm combined adjacency information entropy and adjacency centrality based on multi-dimensional attribute mapping and fusion.The algorithm used weighted adjacency information entropy as the amount of information of neighbor nodes,used adjacency centrality to measure the importance of a node in propagating information about its neighbor nodes,and identified key structural hole nodes in the network by representing the local attributes of structural hole nodes as the ability of nodes to propagate information.Experimental results show that,compared with existing methods,under datasets with different network scales and network structures,the total scores of the three evaluation indicators ofξ,τand network average information entropy are 0.470,1.679,and 4.027,respectively,which are all optimal.It shows that the algorithm has more superior and stable performance.Moreover,the algorithm still has a low time cost when applied to large-scale networks.

关 键 词：结构洞 多维属性融合 信息传播能力 邻接信息熵 邻接中心性 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]