关于复杂网络节点的加权融合感知分类算法  

Weighted Fusion Perceptual Classification Algorithm for Complex Network Nodes

作　　者：李献军 孙科学[2] 张少芳 王月春 LI Xian-jun;SUN Ke-xue;ZHANG Shao-fang;WANG Yue-chun(National Engineering Lab for DBR,Shijiazhuang Hebei 050021,China;Nanjing University of Posts and Telecommunications,Nanjing Jiangsu 210023,China)

机构地区：[1]灾备技术国家工程实验室,河北石家庄050021 [2]南京邮电大学,江苏南京210023

出　　处：《计算机仿真》2020年第8期224-227,270,共5页Computer Simulation

基　　金：河北省高等学校科学研究项目(Z2019060)。

摘　　要：对复杂网络节点进行合理有效的感知分类,有利于改善网络运行的可靠性,增强对关键节点的保护。现有方法大多采取单一指标评估节点,在网络产生波动时易出现全局或者局部影响,导致感知性能存在局限,为此提出了加权融合感知分类方法。方法首先在复杂网络节点无向图模型基础上,针对单一评价指标可能出现的弊端,分析了节点度、抗破坏能力,以及介数三种指标,并改进了连通度算法;然后将节点指标采取层次标记,利用初始判断矩阵计算出加权,并构造节点的评价矩阵;最后根据紧密度公式计算所有节点的近似度,从而实现对节点的感知分类。通过仿真证明加权融合感知分类方法能够有效提高复杂网络节点的感知准确性与高效性,具有良好的鲁棒性。The reasonable and effective perceptual classification of complex network nodes can improve the reliability of network operation and enhance the protection of key nodes.Most of the existing methods use a single index to evaluate nodes,which is easy to have global or local impact when the network fluctuates,leading to the limitations of perception performance.Therefore,a weighted fusion perceptual classification method is proposed.Firstly,based on the undirected graph model of complex network nodes,aiming at the possible disadvantages of a single evaluation index,this method was used to analyze three indexes:node degree,anti-destructive ability,and intermediate number,and to improve the connectivity algorithm.Then,the index of nodes was marked with hierarchy,weighted through initial judgment matrix,and the evaluation matrix of nodes was constructed.Finally,the approximation degree of all nodes was calculated according to the compactness formula,so as to realize the perceptual classification of nodes.The simulation results show that the weighted fusion perceptual classification method can effectively improve the accuracy and efficiency of complex network nodes,and has good robustness.

关 键 词：复杂网络 节点感知 加权计算 融合指标 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]