Analysis Of Centrality Method For Dynamic Networks

There often exist a small number of key nodes which are important for the complex system in large-scale complex networks. Therefore, the work that analyzing the key nodes or the centrality in the complex network is practical undoubtedly.Currently there exist many centrality measures which mainly apply to static networks. However, the complex system changes over time in nature. As we can see, the analysis of static networks will lose much information. This paper addresses how to find the key nodes which contain a lot of information in the network by studying three static centrality methods. In order to measure the importance of nodes and key nodes changes in the dynamic network. Four dynamic centrality methods which combined with the dynamic characteristics of the network are designed. To further illustrate the application of the dynamic centrality, a novel community core mining algorithm based on the dynamic centrality is constructed for detecting the substructures which contain most information of the dynamic network during a period of time.We demonstrate the static and dynamic version of the centrality methods on a real Scientists cooperation network. The experimental results demonstrate the different characteristics of four dynamic methods. In addition, because there is no uniform evaluation in dynamic centrality, to handle this problem we develop a centrality evaluation called CenEvaluate to compare and analysis the four methods reasonably, which leading to some interesting results.