Bi-graph random walk sampling of directed online social networks

Abstract

During the past several decades, various Online Social Networks (OSNs) have experienced a huge development and have given rise to many related studies. However, the huge size of OSNs brings many difficulties to researchers and kinds of sampling methods have been used to get a relatively small but representative sample. Although some traditional sampling methods (e.g. Random Walk and its several improved forms) can help us to get a high-quality sample, there is still a lot of space for improvement in convergence rate, efficiency and their performance in directed OSNs. In this paper, we focus on sampling directed OSNs and propose Bi-graph Random Walk sampling (BRW) as a new sampling method. During each iteration of sampling, we treat directed structure as a combined graph of in-graph and out-graph, and then use a two-stage procedure to sample network. By evaluating this method in both synthetic graph and real OSNs, we find that BRW can achieve higher efficiency and faster convergence rate than traditional sampling methods.