Research On Community Mining And Hierarchical Structure Detection Method In Communication Network

Public communications network is the critical infrastructure for our country and also is theimportant route for people to exchange information. The network relationships reflect not only thecommunication between users, but also reflect the social relations between people, socommunications network relationships mining is an important means to analysis the socialrelations. It is found that there are community structure and hierarchical community structure inthe communications network. The two structures are the important manifestation of socialrelations.It is an important way to analysis the community structure by regarding the communicationsnetwork as a complex network, but due to the size and complexity of the communicationsnetwork, the existing community detection algorithms may have the high computationalcomplexity or detect the low quality communities, and for the existing hierarchical communitydetection methods, we find that the multi-resolution methods which are researched more andworking better, can not detect the natural hierarchical structure in the network, they give theresults of multi-resolution. Therefore this paper depends on the program which is the key projectin the field of national “twelve five-year”863plan, studies the efficient methods for detecting thecommunity structure in communications network. First, we study an accurate and fast communitydetection method in complex network, to find the community structure in large networkefficiently. Second, we study a method can detect the hierarchical community structure directly.Main work and contributions of this paper are outlined as follows:A community detecting algorithm based on transferring the similarity is proposed. Byanalyzing the existing similarity measure methods, we find the local similarity measure methodscan not compute the similarity between the nodes whose distance is more than2. And the globalsimilarity measure methods may fail to compute the similarity between some nodes in thenetwork. Therefore, we propose a similarity measure method based on transferring the similarityto accurately compute the similarity between the specified node with the other nodes in thenetwork. Our algorithm firstly detects the suspected core nodes as seed nodes which may belongto the potential communities, then uses our similarity measure method to compute the similaritybetween the seed nodes with the other nodes, lastly set the nodes in the community which theirmost similar seed node is in.We test our algorithm in different networks, the experiments resultshow our algorithm is effective.A hierarchical community structure detecting algorithm based on local modularity isproposed.When the community grows based on the local modularity, the value of local modularityvaries with the closeness of the community.This paper defines the discrimination of hierarchicalstructure by the community size when it produces the maxima and minima value of the localmodularity.Then determine whether the growing community arrives the boundary of thehierarchical structure or not. We test our algorithm in different networks and compare with themulti-resolution algorithm. The experiments result show our algorithm can detect the hierarchicalcommunity structure more accurate and directly. Based on the above researching work, and combined needs of the project,we design asystem to detect the community structure in the communications network. The preprocessingmodule filters and transforms the communications network data, the community detection moduledetects the community and hierarchical community structure in the transformed data. We use twolarge actual communications network data to test the system in the simulation condition lastly.The experiments results verify the usability of our system, and further verify the effectiveness ofour algorithms.