Estimating average closeness centrality of a large graph in python in a reasonable time

Question

I have a large network file (about 7 million nodes and 7 million edges) that i've imported in a networkx graph. I need to estimate its closeness centrality average across all nodes. I've tried implementing the Eppstein-Wang algorithm (which computes the Single-Source-Shortest-Path for a random node k times to produce this estimate). However, for a graph this size this computation needs to happen about 1000 times to produce a sufficiently precise result, which would take me about 10 hours on my machine.

E-W has a complexity of theta(k * m) (where m is the number of nodes). Are there any other estimation algorithms that have less computational complexity and could get me an acceptable result within, say, around one hour?