Research On Community Detection Algorithm By Using Communication Information

With the rapid development of technology and economy, mobile phones almost become the necessary communication tools for everyone. Therefore, mobile phone user group constitutes a huge mobile communication network. Voice calls or text messages are often used to communicate for mobile phone users. Then group relations in the network can be found by analyzing the relationship between mobile phone users. The community detection method is mostly used in the current study on group relation both in complex network and similarly, in mobile communication network. The community detection method can extract the close group in mobile communication network and recommend personalized services for these groups. Due to the large scale of mobile communication network, it is necessary to do the parallel research on the community detection. In view of the above description, mis paper involves three aspects of the study as follows:(1)The close subgroup of mobile phone users detection algorithm in the mobile communication network is proposed.Due to the problem of classifying relationship circle of a single user, the close subgroup of mobile phone user detection algorithm is proposed. Firstly, the mobile social network is constructed according to mobile phone users’communication behavior. Secondly, the definition of triangle structure in mobile network is proposed, and then finds out all the triangles of the single user according to mobile social network. Finally, the relationship circle of the user is classified by analyzing the reach-relation of triangles. This algorithm can not only detect the close subgroup of users, but also find the change of users close subgroup, which provides a method for researching the dynamic change of community.(2) The community detection algorithm based on core subgroup in the mobile communication network is proposed.Due to the problem of selecting the seed randomly of LFM algorithm, a community detection algorithm based on the core subgroup, called CSCD algorithm, is proposed in this paper. Firstly, the core subgroup among all the subgroups is found by close subgroup detection algorithm. Secondly, the definition of the connectivity between subgroup is proposed, and then expends the communities according to the the connectivity threshold condition. Finally, the definition of overlap degree between communities is proposed, and then the result of communities is further optimized according to the overlap degree threshold condition.(3) The parallel community detection algorithm based on Spark is proposed.Due to the problem of taking too time on community detection in large-scale complex networks, this paper uses Spark for implementing parallel CSCD algorithm-PCSCD algorithm.Most current community detection method is based on node. This paper finds out the community structure by extracting step’node-triangle-subgroup-community’ step by step. Compared with the method basing on node, the one basing on triangle structures in this paper consider not only nodes but only edges. In that way, the group connectivity can be tightened and their structure can be stabilized. The efficiency and feasibility of CSCD algorithm for detecting communities in mobile social network are verified by an empirical evaluation of the method.