Research On Community Detection Algorithm In Social Networks

Community structure, as an important feature in social networks, represents a set ofspecific objects. It can reflect the heterogeneity of social networks to a certain extent.Detecting community structure in social networks can help people to have a deeperunderstanding of the internal regulation of specific groups in social networks. As a result,we can develop the potential value of these groups. In this paper, we mainly make aresearch on community detection algorithms in the social network analysis, and propose anew solution for the existing problems of some community detection algorithms.Firstly, having studied the basic characteristics of social networks and analysedtopology similarity between nodes in the networks, we found that the existing relatedalgorithms ignore the similarity between nodes with close indirect relationship when theycompute node similarity. To solve this problem, we propose a new community detectionalgorithm based on node similarity AINS. The algorithm takes full account of the directand indirect relationships between nodes, with a better community classification result.Secondly, having studied the existing community detection algorithms, we found thatsome community detection algorithms must depend on priori knowledge such as thenumber of communities, the size of community, etc, but they can not guarantee that thepriori knowledge meets the real situation of the network with an unreasonable communitystructure. Meantime, the core nodes in the networks often control the stability and overalldevelopment trend of community they belong to, and play an irreplaceable role in thecommunity. To solve this problem, we propose a new community detection algorithmICNBC, which is based on the importance of the core nodes. The ICNBC algorithm doesnot depend on the number of communities as a priori knowledge, and takes advantage ofthe node similarity to complete the initial division of core nodes, then we can get thereasonable number of communities in the networks and finish detecting communitystructure.Finally, we verify the performance of AINS algorithm and ICNBC algorithm onartificial networks and real networks. The experiments proved that the two algorithms canachieve better results of community detecting.