Performance comparison of sampling techniques for web-based networks

Abstract

The gaining popularity of social networks is attracting a large number of researchers to study the behaviour and characteristics of social networks at a large scale. But it is difficult to capture full view of many social networks due to their large size and access limitations. Therefore, sampling techniques are essential to analyse Online Social Network's characteristics and behaviours. It is a challenging task to create a small but representative sample from a large social graph having millions of nodes. Many graph sampling algorithms have been proposed in past like BFS (Breadth First Search), DFS (Depth First Search), Snowball sampling, Random Walk sampling and their variations. In this paper, we evaluated the performance of Random Walk (RW) and Metropolis Hastings Random Walk (MHRW) algorithm on web-based network datasets. Evaluation is done on the basis of two parameters: average path length and average clustering coefficient. Our results show that MHRW technique performed better than RW technique for both the parameters.