An Incremental Community Detection Algorithm And Application

Complex networks theory has been widely used in analysis of complex systems in many fields such as social,biological science.Community detection is a significant method for identifying potential properties of one complex network.For the application on dynamic networks which are special and common,a few incremental modularity-based algorithms have been proposed recently.As far as known,all of the existing incremental methods takes the evolution of networks as a sequence of edge-grain increments and are very sensitive to the sequence.However,in real world,a large quantity of networks change in node-grained,in which classical methods may lead to poor performance.To address this problem,we proposed an incremental modularity-based community detection method after much research and practice.The results of experiments on a few real-world networks shown the superior performance in terms of community quality and time consuming.The clustering accuracy was also shown by comparison between real community structure and that got by our algorithm on part of networks.Besides,the experiments also indicated that node-grained methods differ from edge-grained methods thoroughly.We also applied our algorithm to Web Service clustering.We organized a similarity-based network based on semantic annotations,document clustering and manifold learning.Then we used our algorithm to detect communities.The applicability was proved to be excellent by our case study which indicated that one community is closely related to one Web Service functional cluster.