| Because of the inherent heterogeneity of real networks,there are great differences in the role of different nodes in network structure and function.Important nodes refer to a special kind of nodes that can influence the network structure and function to a greater extent.The measurement results of node importance are different under different evaluation criteria.From the point of view of network propagation dynamics,the spreading influence of a node depends on the final average diffusion range initiated by the node.From the perspective of network robustness and vulnerability,whether a node is important is determined by the extent to which the network connectivity or network efficiency is affected by the removal of the node.Accurate mining of such critical nodes in the network is of practical significance to control the spread of information,to suppress the spread of the epidemic,to accurately launch product advertisements,to find the important pathogenic genes and so on.Based on the characteristics of network topology,this paper studies the importance ordering of single nodes in complex networks from the perspective of information propagation and network robustness.Furthermore,the optimal effect of multi-propagation source combination is studied.1.This paper proposes an important evaluation algorithm for complex network nodes based on neighborhood similarity.In order to prevent the risk that network function can paralyze,researchers have proposed a number of ways to investigate the effects of node removal or contraction on network structure and function which can be used to guide the construction of a new system with more robust structure.In this paper,we propose a new local metric which only needs to obtain the neighborhood information within two hops of the node to rank node importance.Firstly,we calculate the similarity of node neighbors by quantifying the overlap of their topological structures with Jaccard index;secondly,the similarity between pairs of neighbor nodes is calculated synthetically,and the redundancy of the local link of nodes is obtained.Finally,by reducing the influence of densely local links on ranking node importance,a new local index named LLS that considers both neighborhood similarity and node degree is proposed.The results of this research are of practical significance for the characterization of the destruction and structural reliability of large-scale networks.2.A node influence ranking algorithm based on node's coreness number and ‘structure hole' feature is designed.In the well-known coreness centrality measure,nodes' influence is ranked only by summing all neighbors'-shell values while the effect ofthe topological connections among them on the nodes' spreading ability are ignored.In this paper,network constraint coefficient is introduced to measure the constraint of the node forming structure hole.A node with a small constraint coefficient indicates that the degree of the node is large and the connections among neighbors are sparse.By integrating the node's constraint coefficient which can comprehensively evaluate the number and the topological connections of the neighbors of the node,we construct a correction function to renew its neighbors'-shell values.The ranking effect of different algorithms compared in multiple real networks showed that our methods work more effectively in ranking the most influential nodes than other measures.Our results could shed some light on how to utilize the network constraint coefficient to decrease the effect of highly clustered nodes on the nodes' spreading ability.3.A node ranking method based on information transmission rate is proposed.At present,many classic sorting indicators based on the structure of the network,such as degree,semilocal,betweenness and closeness centrality have been proposed to evaluate node spreading influence.However,these indicators ignore the key factors that determine the effectiveness of information dissemination: the probability of transmission.Accually,the probability of transmission of different content in an online network may vary greatly.Research has shown that degree and semilocal centrality can obtain better influence measurement result when information transmission rate is small,and betweenness and closeness centrality is usually better when information transmission rate is larger.The sensitivity of these classical node influence sorting algorithms to the information transmission rate indicates that: the ranking method should give different influence ranking results under different transmission probability.Consider the effective accessibility path and information transmission rate between nodes and three-step neighbors,a node influence ranking algorithm ASP is proposed in this paper.The results on many real networks and artificial networks showd that the proposed ASP index can rank nodes influence more accurately,and is less sensitive to the rate of information transmission when compared with degree,semilocal,betweenness centrality,closeness centrality and SP index.4.An efficient algorithm for Mining a set of influential spreaders in Complex Networks is proposed.Notice that the network clustering coefficients will increase with the removal of peripheral nodes by the K-shell decomposition method,we select nodes with the highest k-shell value and interconnected with each other as the core to form the primitive subgroup.Then for each neighbor node of the primitive subgroup,when the number of connections between the node and the nodes in the subgroup is no less than the number of its remaining edges,the node would be added to the subgroup.The cluster expanded by this way is densely connected and finally the nodes with largest degree in each cluster form a set of influential spreaders.The experimental results of the proposed algorithm in six real networks and LFR artificial networks indicate that our method can identify the influential spreaders more accurately than the ones generated by the discount degree method,VoteRank,LIR,k-shell and degree centralities. |
PDF Full Text Request