Research On Community Detection In Complex Networks

With the popularization of information technology,networks have become an integral part of modern life.Many complex systems can be modeled as network with nodes for individuals and edges for relationships between them,permeating field of social,biological,and computer and so on.Complex network science provides great convenience for us to research these systems.As the research continues,the structure complexity,function and size of the complex networks we have known are becoming larger and larger.Studying complex networks by the network topology directly is becoming more and more impractical.Community structure is one of the ubiquitous and significant topology characteristics of complex network.It can help us to study the structure and functions of a complex network.Detecting the community structure can help us reveal the topology of complex networks,analyze their function and predict their actions.The research on community structure detecting has nerve been stopped since Girvan and Newman defined the community structure is the division of network nodes into groups within which the network connections are dense,but between which they are sparser.In this thesis,we investigate several kinds of classic community discovery algorithm and analyze their performance briefly from global and local two aspects at first.Based on this we propose two community structure detection algorithms in this thesis.The process of community detecting is also a kind of nodes clustering,which nodes in the complex networks have common feature means they belong the same community.Based on the theory of information transmission,the nodes are abstracted from a complex network into a multi-dimension data set.Combined with the traditional multi-dimension data clustering algorithm,we propose a new community detection algorithm.Compared with the traditional global community detection algorithm,the new algorithm can get a higher precision while maintaining the same time complexity.With the coming of the era of big data,the size of complex networks is poised for explosive growth.Meanwhile the topology of a network may evolve over time.Detecting community structure form those dynamic complex networks quickly became a high-profile topic and formed a new important research direction.Based on the traditional density-based algorithms and local community structure detection algorithms,we propose a new dynamic overlapping community detection algorithm.It can detect overlapping community structure from dynamic complex on the basis of the past community structure.