Research For Large Mobile Phone Calling Networks Based On Community Detection

Large scale mobile communication,network reveal the pattern of social operation.It is one of the most significant research in complex network.In this paper,we use the community detection algorithms to carry on the research of the social network of large scale mobile data.The specific work of this paper is as follows:1.The structure of call/SMS,call/SMS network is revealed.The phone network structure includes internal calls,outside the city to internal calls,inside the city to outside calls,iternal SMSs,outside the city to internal SMSs,inside the city to outside SMSs.And studying the research methods of network structure,such as the distribution of nodes,the edge distribution,the distribution of nodes’ degrees,the distribution of nodes’ strength,and the correlation of the network structure.2.Comparing the performance of the three algorithms in the evaluation index of the community division in the process of large scale real call network.Using real mobile communication data of the first week in that month,building the mobile phone call network(MPC).The community detection method(CNMF),InfoMap(F-measure),and LPA are used to evaluate the association of MPC network.The performance of the proposed algorithm is evaluated by using the density,clustering coefficient,conductivity,triangle participation ratio and accuracy,recall,F-measure,accuracy and NMI.3.Comparing the performance of the three algorithms in a variety of network characteristics when dealing with large scale real call network.The definition of a variety of network characteristics is proposed: community size,community level,community level,community density,community average friends,community center node importance.By computing the CNMF algorithm,InfoMap algorithm and LPA algorithm to the community of MPC network results of the community characteristics of the three algorithms are compared.CNMF,InfoMap and LPA three algorithms are used to carry out the empirical research on the actual network.The CNMF algorithm has high complexity,but be stable and structural strength.InfoMap algorithm has strong entirety and fast,but the clustering effect is poor,when the community structure of network is weak.LPA algorithm is called many times in this paper to achieve local optimum in the network.The calculation of evaluation is better,but the entirety is weak.