The Research Of A Dynamic Community Detection And Evolution Method In Opportunistic Networks

In complex networks the community feature is an important characteristic.As a special form of complex networks,opportunistic network also has a feature that some similar nodes get together and presents the community structure.Due to the opportunistic network that based on node meet opportunities to communicate,so the topology of opportunistic network is continuously changing and the community structure is also in constant changing as the change of network topology.To study the dynamic community can help make a better understanding of the network structure and make better use of network,To address this problem,this paper made the following research work:1、The traditional community detection algorithms in opportunistic networks generally lack of comprehensive consideration of social connections between nodes,relationship strength and the intimacy.Aiming at this issue,this paper put forward a kind of community detection method based on intimacy(CDMI).This algorithm firstly computed the value of social pressure indicator and relationship strength based on the single cycle meeting history information between nodes.so as to determine which nodes in the network were connected with edge inside the corresponding period.Then calculate the intimacy between nodes and node with the community.According to the gathered coefficient,seed node information could be obtained and community structure detection could be completed.Simulation results is compared with node dynamic belonging algorithm(NBDE),it verifies the feasibility and accuracy of the method.In addition,the method also can get overlapping community structure.2、In opportunistic network,node belongingness is one of the most important aspects of community research,a method that could judge which community a new node belonged to was proposed.This method mainly treated meeting frequency、 meeting duration and the number of meeting that between two nodes as the input vector of neural network.The weights and thresholds of the model were adjusted constantly in the process of training this model.After the model was completed,a vector that was made up of new nodes was fed into the model and through network computing winning neuron could be obtained,the winning neuron was representative of the input data classification,based on this new node belongingness could be judged.Through making tests on the artificial data sets of LFK,the result shows that this method can effectively determine which community new nodes belong to.