An efficient algorithm for sampling of a single large graph

Graph Databases offer a very influential way to provide an instinctual representation for many applications spanning from social networks, web networks to biological networks. In the current era of big data, the size of the graph is increasing exponentially. It is difficult for the conventional machines to analyze a whole graph. To overcome this, the characteristics of the large graphs are estimated via sampling in order to identify trends and patterns in the large graph. The existing sampling techniques such as random node and random walk do not provide consistent efficiency over the graphs. In this paper, an efficient sampling algorithm named Influence sampling (IS) is proposed which sample the graphs by analyzing the degree of the vertices of the graph such that the most influential vertices remain in the graph sample. The experiments are performed over three real life datasets and the performance is compared with the three existing sampling algorithms. It is shown that IS performs well in the terms of accuracy.