Font Size: a A A

Research On Mining Algorithm Of Key Information In Complex Networks

Posted on:2021-10-12Degree:MasterType:Thesis
Country:ChinaCandidate:J L ZhangFull Text:PDF
GTID:2480306110495084Subject:Control Engineering
Abstract/Summary:PDF Full Text Request
Complex networks are closely related to all aspects of human life.It is essential to identify key node clusters for mining useful information in complex networks.Key node clusters can control the entire network,which is of strategic importance to suppress the spread of the epidemic,accelerate information dissemination,and promote new products.Based on the complex networks theory,this paper studies a variety of algorithms for identifying key nodes and key node clusters in complex networks,and proposes two new methods for identifying key node clusters to avoid the phenomenon of "twin nodes" and "canceling interference nodes".The contents are listed as follows:(1)To avoid the phenomenon of "twin nodes",this paper proposes an algorithm to identify multiply nodes,the maximum connected component decomposition method(MCCD)based on maximum connected component decomposition.The algorithm comprehensively considers the different topological properties of nodes and combines them with the largest connected component decomposition.MCCD can identify key nodes that have a greater impact on the network but are not the most important.Therefore,the key node clusters identified by MCCD are less obstructive to mining.Finally,an experimental study of the susceptibility-infection-recovery(SIR)infectious disease model is carried out in four real networks to verify the performance of the proposed algorithm and seven centrality-based and heuristic algorithms.Experiment results show that the proposed algorithm has excellent performance with respect to the propagation speed,the propagation range and the distribution range of the identified key node clusters,and effectively avoids the phenomenon of "twin nodes".(2)To avoid the phenomenon of "destructive interference nodes",this paper proposes an algorithm to identify a group of key nodes,generalized degree discount heuristic and K-Shell(GDDKS)based on generalized discount degree and K-Shell.The algorithm proposes the concept of generalized discount degree.Using this theory,we propose the concept of degree of a group of nodes,and combine the generalized discount degree with K-Shell algorithm to identify a group of key nodes.the overall influence of key node clusters is considered in GDDKS,which can identify a group of key node clusters that have a large impact on the network and are widely distributed.In addition,GDDKS can identify a group of key nodes that have a large impact on the network and are widely distributed.Finally,an experimental study of the susceptibility-infection-recovery(SIR)infectious disease model is conducted in 6 real networks,and the performance of the proposed algorithm and 10 other algorithms based on centrality and heuristics is analyzed.Experiment results show that the proposed algorithm outperforms other algorithms regarding the propagation speed,the propagation range and the distribution range of identified key node clusters,and can effectively avoid the phenomenon of "destructive interference nodes".Based on a complex networks,this paper adopts the largest component decomposition and generalized discount degree to improve the key node cluster recognition algorithms.The proposed algorithms solved the problems of "twin nodes" and "destructive interference nodes",which can provide some insights into identifying key node clusters and thus has theoretical and practical significance.
Keywords/Search Tags:Complex networks, Key node clusters, Propagation dynamics, Maximum connected component decomposition, Generalized discount degree
PDF Full Text Request
Related items