An Incremental Algorithm for Estimating Average Clustering Coefficient Based on Random Walk

Clustering coefficient is an important measure in social network analysis, community detection and many other applications. However, it is expensive to compute clustering coefficient for the real-world networks, because many networks, such as Facebook and Twitter, are usually large and evolving continuously. Aiming to improve the performance of clustering coefficient computation for the large and evolving networks, we propose an incremental algorithm based on random walk model. The proposed algorithm stores previous random walk path and updates the the average clustering coefficient estimation through reconstructing partial path in an incremental approach, instead of recomputing clustering coefficient from scratch as long as graph changes. Theoretical analysis suggests that the proposed algorithm improves the performance of clustering coefficient estimation for dynamic graphs effectively without sacrificing in accuracy. Extensive experiments on some real-world graphs also demonstrate that the proposed algorithm reduces the running time significantly comparing with a stateof-art algorithm based on random walk.