基于多属性偏好信息集结的复杂网络重要节点辨识

Identify important nodes in complex network based on aggregation of multi-attribute preference information

为了准确辨识复杂网络中的重要节点,避免单一属性指标评价节点重要性出现的偏差,提出了一种基于多属性偏好信息集结的复杂网络重要节点辨识方法。首先根据节点的局部特性、全局特性及空间位置等特性,选取度中心性、介数中心性、紧密度、结构洞、K-核(Ks)五个属性指标构建多属性复杂网络重要节点辨识模型,对节点属性偏好信息进行分析、集结和融合;然后将网络中所有节点作为评价主体,构建复杂网络多属性决策矩阵,根据熵理论对节点属性赋权,计算其与理想重要节点的贴近度,对节点重要性进行精细化排序。将该模型应用到"风筝网络"和"ARPA网络"中,根据节点重要性辨识结果对网络进行破坏性实验,结果表明,该方法的准确性比已有方法更高。

A method of identifying important nodes in complex network based on aggregation of multi-attribute preference information was proposed for purpose offacilitating identification of important nodes in complex network accurately and avoiding deviation of node importance evaluation based on single attribute index.Firstly,a multi-attribute model for identifying important nodes in complex network consisting of degree centrality,betweeness centrality,closeness centrality,structural hole and K-core indicators was built based on the local characteristics,global characteristics and spatial location of nodes,for analyzing,aggregating and fusing node attribute preference information.Secondly,a multi-attribute decision matrix for complex network was established with all nodes in the network as evaluation subjects,and the node attributes were weighted in accordance with the entropy theory,to calculate its closeness to the desired important node,and sort the importance of network nodes in a refined way.Finally,the model was applied to the"kite network"and"ARPA network",and destructive experiment of the network was conducted according to the result of node importance identification.The results suggest that the method is superior to the existing methods in respect of accuracy.