Skip to Main Content
With the rapid development of large-scale online social network applications, understanding the community characteristics of online social networks is benefit to improve application performance. Characteristics of communities are studied based on real measurements of You Tube, a popular online social network. Adapt to the huge scale of online social networks, we modify the original community discovery algorithm, the label propagation algorithm, to reach a balance between goodness of community division and time efficiency. We analyze the distribution of community and group sizes and the relation between community structure and group membership user explicitly claimed. Our experiment show that both the community sizes and group sizes follow a power-law distribution and the dependency of community membership on the group membership is evident, but the latter is neither the only nor the main origin of the community structure. Also, we studied assortative matching, degree distribution of the entire network and the largest scale of community. The results confirm that assortativity matching of the entire network is much higher than inter-communities, both in degree and out degree distributions of the largest scale community and the network satisfy power-law property.