Compressing big graph data: A relative node importance approach

Given a complex graph of large network, people always want to capture its desirable information of interest. In order to dig the underlying structure of the extremely large graph dataset, compression is necessary, which also helps in further communication and computation. In this paper, by invoking a newly proposed node centrality metric named relative node importance (RNI), an effective compressing scheme is presented for complex graph datasets. Besides measuring the distance and connectivity distribution of the graph structure, we firstly take k-core distribution into consideration. Compared with the existing schemes, the proposed one has lower computational complexity and fits different kinds of networks, e.g., social network, the World Wide Web (WWW) and autonomous systems (AS) network. Numerical results show that our RNI-based method outperforms other schemes and well preserving the basic features of a graph.