Research On The Interdependent Relationship And Community Detection Algorithm In Mobile Social Networks

In this paper we have analysis the structure characteristics and network characteristics of the Mobile Social Network. According to complex network theory, we proposed a group growing scale-free network model which can reflect the Mobile Social Network. We construct two kinds of interdependent networks because the users and mobile devices in the Mobile Social Network have different dependent relationships. One is interdependent networks with bidirectional inter-edges that can reflect the one-to-one relationship of the Mobile Social Network, the other is interdependent networks with directional inter-edges that can reflect the multiple support-dependence relationship of the Mobile Social Network.In this paper we adopt random attack and target attack strategies to attack the Mobile Social Network. Then we observe the change of the netwoks with different inter-relationships, and analysis the robustness of the networks respectively. Through theoretical derivation we get the threshold which can cause the cascading failure of the networks. Then we use computer to simulate the process of cascading failure in the networks. When we remove certain fraction of nodes the network will split to some subsets, and the remaining nodes which still are functional form the remaining largest clustering. The size of the remaining largest clustering can reflect the robustness of the networks. When the removal fraction of nodes reach the threshold, “avalanche” occurs in the network, it means that there is not exist remaining largest clustering after cascading failure. The results prove that the Mobile Social Network with interdependent relationships are more frigile than the single network under the random attack, but it is opposite under target attack; Whether under the random attack or deliberate attack, the Mobile Social Network with multiple support-dependence relationship is more robust than the Mobile Social Network with one-to-one relationship.According to the dynamics of the real network, we develop an adaptive community detection algorithm, which can use in dynamic networks. Different from the traditional community detection algorithm, the method of adaptive is introduced in this algorithm. It is not necessary to consider the entire network topology, the community structure can be divided according the past community partition and the change of the network. In this paper we analyze the accuracy of this algorithm through executing the algorithm in the real datasets and s ynthetic datasets. The researches show that, when we adopt the NMI evaluation criterion, the algorithm has a better performance compared with other algorithms.Finally, we made a summary for our research work, anlysis the deficiencies existed in the research, and put forward the future research prospects.