Research On Comprehensive Evaluation Of Network Node Importance Based On Multiple Indicators

With the discovery of the small-world and scale-free characteristics of complex networks,the study of complex networks has become a hot research field.The actual network contains many nodes,but due to the heterogeneity of the network,different nodes play different roles in the network.It is not only of theoretical significance to comprehensively evaluate the importance of complex network nodes and explore the maximization of network influence,but also of great application value in many fields,such as epidemic control,advertising,communication network security,prediction of hot research results and protein interaction.The research of this dissertation develops from two levels.Firstly,considering multiple centrality indicators in the network comprehensively,the centrality indicators are weighted reasonably from the subjective and objective dimensions.VIKOR,the classical method in the multiple attribute decision theory,is introduced into the comprehensive evaluation of network nodes.Secondly,for a given complex network and its information dissemination model,how to find the initial node set for information dissemination,to maximize the number of nodes was eventually influence,we take into account the two dimensions of importance and dispersion of nodes.H-index is used to describe the importance of nodes,and the minimum distance is used to describe the dispersion of nodes.An extended clustering method is proposed to select the important Top-k nodes.The main work is as follows:1.The combination weighting VIKOR(CW-VIKOR)method is proposed to evaluate the importance of nodes.Weighting is the key step of multiple indicators evaluation method and has important influence on identifying important nodes.However,the existing evaluation methods of multiple indicators important nodes mostly consider the indicator weight from a single dimension,which has certain limitations.Considering the two dimensions of subjective and objective,this dissertation puts forward a weight optimization strategy.The VIKOR method is a common method of multiple attribute decision making.It takes full consideration of the subjective preference of the decision-maker,and applies it to the complex network for the first time,and proposes the combination weighting VIKOR method.Experiments show that CW-VIKOR method can effectively identify important nodes in network.2.An extended clustering method is proposed to select the important Top-k nodes(HD-Cluster).Aiming at the important and scattered properties of Top-k nodes,this dissertation adopts the idea of clustering to divide the nodes into different clusters,and takes the center of each cluster as the key node of Top-k.Each cluster is led by a center node with little overlap between the different clusters.Since clustering method is sensitive to the selection of initial center,we propose an initial center optimization strategy based on node influence and dispersion to select the initial seeds reasonably.Experiments in four different types of networks prove that HD-Cluster method can effectively identify multiple key nodes in the network.In this dissertation,the importance of network nodes is comprehensively evaluated from the perspective of multiple indicators,and CW-VIKOR method and Top-k important node identification method are proposed.Compared with the existing methods,CW-VIKOR method and HD-Cluster method have better performance.This research has important theoretical significance in mining network information and can be well applied to the actual network,which has potential application value.