| This thesis discusses two research topics in the domain of complex networks,with an emphasis on identifying key spreaders from the perspective of nodes,as well as determining their communities from the perspective of groups of nodes.Specifically,it aims to identify those nodes that are more influential in the entire network,as it is applicable across a wide range of domains,for example,making information or advertisements viral over social media networks,or controlling rumor or epidemic spreading in a network.It has always been one of the fundamental topics of research to detect communities.By leveraging graph topological structure,community detection focuses on node high-order closeness.Similar edges or nodes can be grouped together in the same community while dissimilar ones can be separated into separate communities to reveal the graph structure in a coarser resolution.There have been various incentive models/methods developed and deployed to address the importance of nodes and community detection.Although their exceptional efficacy,these approaches have some inherent limitations that must be addressed.Therefore,in this dissertation study,we suggest several approaches for detecting node importance and community detection.Following is a list of contributions and innovations:1.Assessing the importance of nodes in complex networks is crucial for investigating the survival and robustness of networks.There are numerous potential applications,such as preventing outbreaks,spreading viruses on computer networks,viral marketing,sickness spreading,and so on.These problems are usually unable to be solved by many heuristics with low time complexity.Therefore,a plethora heuristics have been suggested regarding influential nodes identification.A simple example is that degree centrality is the easiest to determine,whereas closeness centrality is more complex and cannot handle large-scale or complex networks.Still,most of the current heuristics have issues,for instance,low rating accuracy or high time complexity.To deal with this issue,on the basis of local information of a network,we propose a novel approach to identify influential nodes called locality-based model(LSM).The proposed LSM takes into account the kshell,degree,and number of triangles in a network.The proposed LSM can effectively identify the importance of nodes with low time complexity.First,lines(connectivity factor)are computed based on the properties of nodes connected to them.Then,each node’s contribution to the importance of lines is computed.Finally,the degree and k-shell of nodes as well as their contribution to the importance of lines are taken into account.To validate the performance of LSM,contrast experiments are performed with benchmarks on different scale-sized networks,where a susceptible infected removed/recovered(SIR)epidemic model is used to check the propagation dynamics of each node.The simulation results under the standard SIR model indicate that the proposed LSM is able to efficiently identify influential nodes in numerous types of networks without giving the parameter setting in advanced.2.By simultaneously using local and global topological aspects,we design a Local-and-Global Centrality(LGC)measuring algorithm.LGC comprises factors such as degree nodes,path distance between nodes,and numerous levels of neighbor’s influence or neighborhood potential.In order to assess the performance of the proposed LGC with respect to the state-ofthe-art methodologies,we performed experiments through LCG,Betweenness(BNC),Closeness(CNC),Gravity(GIC),Page-Rank(PRC),Eigenvector(EVC),Global and Local Structure(GLS),Global Structure Model(GSM),locality-based model(LSM),and Profit-leader(PLC)methods on differently sized real-world networks.Our experiments disclose that LGC outperformed many of the compared techniques.3.By combining neighborhood and path information,we propose an approach called Neighborhood and Path Information-based Centrality(NPIC)to identify the highly influential spreaders in a concern network.To analyze NPIC performance through extensive simulation,we have conducted experiments on real datasets(networks)and applied the epidemic model to verify each nodes efficiency of spreading effect with respect to its surroundings.Simulation experiments on various real datasets demonstrate that NPIC can efficiently identify the persuasive spreaders in the corresponding networks as compared to the existing baseline approaches.4.An approach called global structure model(GSM)is proposed to identify influential nodes in complex networks,taking into account self-influence and concentrating on the global impact of each node.GSM is verified on different real and synthetic networks and compared with recently proposed approaches.From experimental analysis,the proposed approach can efficiently be identifying the nodes influence in a network.5.We propose a novel Relevance-based Information Interaction Model(RIIM)to identify communities in complex networks based on local as well as global topological aspects without providing prior community knowledge and parameter configuration.In order to evaluate the effectiveness of RIIM,we conducted extensive experiments on real and synthetic networks,and the generated performance demonstrates that the proposed approach outperformed the state-of-the-art techniques by effectively identifying the corresponding communities in complex networks.There are 28 figures,30 tables,and 216 references. |
