Towards Unbiased Sampling of Online Social Networks

Download Now Free registration required

Executive Summary

Unbiased sampling of Online Social Networks (OSNs) makes it possible to get accurate statistical properties of large-scale OSNs. However, the most used sampling methods, Breadth-First-Search (BFS) and Greedy, are known to be biased towards high-degree nodes, yielding inaccurate statistical results. To give a general requirement for unbiased sampling, the authors model the crawling process as a Markov Chain and deduce a necessary and sufficient condition, which enables them to design various efficient unbiased sampling methods. To the best of their knowledge, they are among the first to give such a condition.

  • Format: PDF
  • Size: 392.3 KB