Self-similarity of complex networks under centrality-based node removal strategy

Abstract

Real-world networks exhibit complex topological interactions that pose a significant computational challenge to analyses of such networks. Due to limited resources, there is an urgent need to develop dimensionality reduction techniques that can significantly reduce the structural complexity of initial large-scale networks. In this paper, we propose a subgraph extraction method based on the node centrality measure to reduce the size of the initial network topology. Specifically, nodes with smaller centrality value are removed from the initial network to obtain a subgraph with a smaller size. Our results demonstrate that various real-world networks, including power grids, technology, transportation, biology, social, and language networks, exhibit self-similarity behavior during the reduction process. The present results reveal the self-similarity and scale invariance of real-world networks from a different perspective and also provide an effective guide for simplifying the topology of large-scale networks.