The Study Of Methods About Community Detection Based On Local Expansion In Social Networks

Community detection derives from structural analysis of complex networks, such as social networks and biological neural networks, and it plays an important role in many applications, e.g., in the prevention of the virus spread, the strategy of message forwarding and the design of multi-hop ad hoc routing protocol. People’s activities are integrated into network, such as making friends, shopping, learning in online pattern, that motivates the researchers concern on community detection. In view of the large scale and dynamic characteristic in complex social networks, the local community detection methods get more and more attention. Compared with the global methods, not only does the local model have the simple, fast, flexible implementation, but it also has higher application value. This paper adopts the local method to learn the deep internal structure, and proposes two kinds of new local community detection algorithm, which the specific contents are summarized as follows.Firstly, according to the overlap characteristic of the community structure in social network, this paper proposes a local expansion query based method for local overlapping community detection, namely OCLEQ. It uses the query technic to unfold local expansion taking advantage of high flexibility. Furthermore, the algorithm leverages the clique structure and its adjacency, which can realize the overlapping community discovery easily; In addition, it detects and classifies the left nodes through a new metric, to further improve the accuracy of the algorithm.Secondly, according to the characteristic of the edge weight in social network, this paper puts forward a community detection method based on local expansion of the core nodes in weighted networks- WCCE. The algorithm is essentially a local expansion algorithm in layers. The initial step is the expansion on the core nodes,and the next step is the iterative cycle of local expansion and community unification. The process of expansion in layers makes the network scale shrinking, which makes the algorithm simple and feasible. Furthermore, WCCE ensures the quality of community detection through the detection and classification of missing nodes.Finally, this paper validates the above algorithms on benchmark data sets and real data sets. The experimental results show that the community quality and time complexity of the algorithms are obviously improved than the recent works.