The Study Of The Community In Social Network Based On Multiple Index And Its Application

With the development of the large scale network study,it has been a hot spot in thecurrent rearch to discover implicated structure characteristics and to obtain potentialknowledge in the network, which obsesses a great practical value. However, it is difficult fortraditional community detecting method to meet three conditions simultaneously that is lowtime complexity, cluster with high-precision and pior knowledge not required. At present, thescale of network becomes larger and larger with a dynamic change, which makes it hard to getall the information in the network. That means a fast, efficient community detectiong methodis required, which only needs local network information. In addition, people not only concernabout internal structural community and its feature and function, but also become interested tomore valuable details implicated in the network, such as how rumor spread in the network,what kind of role people play in the spread, what kind of position people have in theirinterpersonal relation network, is there any kew person in the criminal network and who is thehead of terroristic organization, etc.There are many measurement indexes for characteristics of the network, but one measureonly partly indicate the characteristics while the node is important and its role is multiple andcomplex, therefore we cannot rely on sigle index. Firstly, based on multiple measurementsindex, this paper explores the limitation of sigle measurement index, multiple indexes relationand distribution characteristics of each index. Secondly, on the basis of above reaserch,multiple node is achieved is by peroperly and effectively selecting degree centrality,betweenness centrality and eigenvector centrality; the importance of node is evaluated byusing the weighted Sum method; BME method is put forward. Then, in line with above threeindexes having a effect on the importance of node in the network, that is index contributes tothe importance of node, vector predication BMIR method for node behavior model is raised.Finally, BIRC method for community delivery is put forward, whichis more accurate andefficient in accordance with local network information. Based on BME and BMIR, thismethod achieves core community by deleting all bridge point and edge between core nodefrom the thought of “from seperation to cohesion”. Then cohesion is realized by completingthe division to the whole network through computing’s attraction on node.