Efficient sampling strategies for large-scale complex networks

In empirical research on large-scale complex networks, sampling is a necessary way to collect data. Current methods commonly-used are Y2H-derived partial sampling strategy and random sampling strategy. Some recent studies have proposed that subnets sampled by these methods may not accurately conserve structural properties of the original network. Therefore, how to improve the accuracy of data collection is raised as a significant problem. We present an effective strategy for sampling in complex networks. The proposed strategy, hub strategy, calls for targeting the highly connected node samples. We demonstrate that in contrast with current sampling methods, hub sampling strategy keeps multiple structural properties of networks more accurately as well as being more economical. Furthermore, we find that when sampling rate decreases, hierarchical modularity is easier to be distorted quantitatively by hub sampling than the other structural properties.