An improved algorithm for community discovery in social networks based on label propagation

Community discovery in social networks can be seen a classification problem for the networks vertices. Therefore, an improved label propagation algorithm for the community discovery is proposed based on the perspective of pattern classification. The algorithm based on secondary classification ideology can be simply described as follows: the social networks are first divided into several original communities based on networks structure and the results of classification are assigned to each vertex of the networks as the label; secondly, the label are spread based on local similarity of vertices; ultimately the vertices which have same labels can be divided into a community. It is a process of secondary classification that can reduce uncertainty of the labels setting and randomness of labels propagation effectively. Experimental results show that the improved algorithm can greatly improve the quality and stability of community discovery.