Real Time Closeness and Betweenness Centrality Calculations on Streaming Network Data

Closeness and betweenness are among the most important metrics in social network analysis. They are essential to the evaluation of various research problems such as viral marketing, network stability and network traffic predictions, which play an important role in social media research. However, both of these metrics are expensive to compute. We propose an efficient online algorithm framework to handle both closeness and betweenness in the situation where network structure changes frequently. Whenever a link change is received as the input, the algorithm utilizes existing facts about the calculation to update centrality values with minimal effort. Experimental results on data sets collected from online social media websites show that our approach is 4-7 orders of magnitude faster for closeness and 2-4 orders of magnitude faster for betweenness calculations over baseline methods. We also show how those two metrics share some common calculations so that the running time can be dramatically reduced when calculated together. To the best of our knowledge, this is the first work to improve the running time when those two algorithms are calculated at the same time on streaming network data.