Looking at the GDS documentation for closeness centrality and it mentions that the score is normalized. My expectation is that the score would range between 0 and 1. However, I am getting results that often exceed 1 and by a significant margin.
My test is based on a graph where I used label propagation with a minimum community size of 10 nodes. The for each of the communities I run a suite of centrality algorithms including closeness.
Looking at the distributions I notice that between 10%-25% of the nodes have closeness scores greater than 1. Is that to be expected?