Local Community Detection Using Content And Links

Community detection originates from the analysis of complex networks, such as social network, and biological protein network. In recent years, with the popularity of information technology development as well as social media technologies, the social networks show a lot of new features, such as, large-scale network, the dynamic nature of information-rich content and other characteristics. Considering a global community detection technology and community detection algorithm that just involves the topology property of complex network mostly have been unable to adapt these new features. Aiming at these new features of the current complex networks, we consider the local community detection technology as research priorities, and combine node attributes and edge structure in the process of detecting the structure of community. The main research achievements are as follows:Firstly, in the process of exploring a complex network core node, we proposed a new method to explore the core node based on semi-local centrality and distance. Semi local centrality can ensure the core network node is an important node. It is different from global centrality. It involved four neighboring nodes when computing a node’s semi-local centrality. The core node of centrality expression is more accurate. Local community detection algorithm can be extended to find a global network of community structures. The distance between core nodes, enables the network core nodes reasonably evenly distributed throughout the network, resulting in high-quality global network of community division.Secondly, in the diffusion process of the local community detection, the content information of complex networks and links information are combined. In recent years, researchers have proposed a variety of complex underlying structure convergence of content and link information network detection algorithm, mostly based approach probability model. The new method proposed is different from the current research hotspot. The links and content similarity function are integrated into the same objective. And then diffusing the local communities through iteration.Finally, integrate these two steps and make experiments with real-world datasets and benchmark datasets showing that the community can get higher quality by this algorithm, and the effect is obvious in terms of time efficiency.