Research On Community Detection Algorithms Based On Complex Network

Complex network is widely present in the human society and the natural world, networkof individuals and the surrounding environment influence each other, interaction, togetherconstitute the complex network. In the real world, complicated network can be seeneverywhere, such as social relations in the scientists cooperation network, epidemic diseasetransmission network, the ecological system of metabolic network, protein interactionnetwork, etc., especially in the information age, since the development of the Internettechnology and the world is getting smaller, and interpersonal contact more closely, thehuman society has gradually evolved into a network of the world. Therefore, the complicatednetwork research has increasingly become a hot research topic.Complex network because of its large scale, with nodes and its complex often has thecertain difficulty for direct study of complex network, and the community as an importantproperty of complex networks, the research communities in complex networks has attractedmore and more attention. Research on community detection in complex networks, is a miningcommunity structure among them, which found its community division; on the other hand isassessed the importance of network nodes, and explore the important node.In recent years, the complex network community detection algorithms research has madeconsiderable development, also appeared many classic community discovery algorithm, suchas GN algorithm, K-L algorithm. But many of these algorithms are required to providecommunity partition number, community size and some prior information, and this a prioriinformation in general is difficult to get. To be able to provide accurate prior information, thealgorithm can accurately community division, but, if cannot get accurate prior information,the results of these algorithms often division are not very satisfactory.Based on the complex network and community discovery and relevant literature, the fullunderstanding of the characteristics of the categories and related algorithm in complexnetwork community discovery, analysis and summary of some of the classic communityfound the advantages and disadvantages of the algorithm. In addition, this paper studies thenode centers in the network and node similarity problem, at the same time, the communitydiscovery algorithm, first proposed the concept of distance centrality, and bring out thedistance from the center of the community discovery algorithm based on (DCCD algorithm)and distance nodes based on the evaluation center.The distance centrality, using the distance between nodes as the foundation， containsthe concept node centrality and similarity. DCCD algorithm is based on the distance of thecenter of nodes as the standard, to select the center node through its center, and then by the similarity to judge other nodes belonging to the community, to complete the networkcommunity division. At the completion of community division, using K-means algorithm toget the community continue to distance from the center of the basis of iterative calculation,recalculate the center node of each community, and then to community division, until thenetwork reached a stable structure. And the distance of the center of evaluation index, notonly consider the node in a network location, more with the connected relationship and degreeof nodes and the direct relationship between adjacent nodes, and the adjacent nodes on thecardiac contribution degree. Therefore, the distance of the center of evaluation index is a morecomprehensive measure.In order to verify the correctness of DCCD algorithm and the distance of the center ofevaluation index, the Karate Club data sets and a series of community discovery algorithmcommonly used classical data set on the experiment, and comparing with other communitydiscovery algorithm results. The experimental results show that, DCCD algorithm inprocessing community center node is more obvious when the data set, can obtain the goodcommunity division, also can discover hidden structures within the community, more highquality community division; the distance of the center of evaluation index can also beimportant node evaluation data better, and compared with other evaluation index, has higheraccuracy.In small static network, community discovery algorithm based on distance centricityevaluation index and the important node has a good community classification results and theevaluation results, but in actual network, more large-scale and dynamic networks. Continueon the complex network community discovery algorithm, therefore, further research andimprovement, to make it better able to apply in actual network, and that is the focus of futurework.