Dynamic Community Discovery And Application Combined With Influence

The study of community structure in social network provide an accurate positioning for the socialization.This study is helpful to the mining of potential commercial value in the later stage.At present,the research areas of community discovery algorithm has expanded to the dynamic network and overlapping community from the static network and non overlapping community.Duing to simple and efficient characteristics,the community discovery algorithm based on the concept of tag communication,which can adapt to the growing social network,has been concerned by researchers.However,the existing improved algorithm is still not very good at solving the instability result problem of the final community division,which is caused by using a random strategy in the stage of label propagation and selection.In view of the research of dynamic network community discovery,if we continue to use the existing algorithm of static network to divide the whole community network,it will cause too much time consuming.In addition,most of the current social network are removed node attributes,but there are still a lot of hidden information available in the network topology are excavated.Those information are used to identify the identity of the node and its community information,that will disclosure of user privacy.Therefore,the following works have been done in this paper:1.In order to mining high-quality overlapping community structures,we proposed an overlapping community discovery algorithm that incorporates the influence of multi-tag propagation.The algorithm selected the maximum number of K bridge nodes to expand the cluster,and obtained several overlapping rough clusters to initialize labels.In labels propagation,we used the asynchronous update strategy,and uesd updated sequences generated by bridge coefficients to update labels.This method improved the quality of the community and the stability of the algorithm.We tested different experiments on real data sets and artificial neural networks.The results showed that the proposed algorithm is effective and feasible.2.At present,most of the social networks have been relatively stable and adjacent time network topology changes smaller.Mining community structures of the whole networks repeatedly costs too much time and is not necessary.In this paper,we propose a dynamic community discovery algorithm.This algorithm obtained the incremental information from the adjacent time network,and got the unadjusted local network by using the community structure of the previous time and the coefficients change of bridge edge.Then this algorithm redivided local community network and got the results of the community division about currernt time.We did experiments to verify algorithm effectiveness by useing the real routing network data and the.artificial network structure topology.3.For the attack model in which the attacker knows the time change of the target node and the community information,we designed a privacy protection algorithm of node commnunity.This algorithm uesd k-community to group all nodes by degrees and community information of nodes.Then this algorithm processes each group anonymously,Finally,we achieved that attackers can not attack the nodes,even if attackers know the target node and community information that we described above.In the process of anonymity,considering the bridge coefficient of the side,we first deal with edges that are not easily change in community structure to ensure the validity of the release graph.After the validation of experiments,the expected effect has been achieved.