Currently {clustree} assumes the user knows the order of resolutions but this might not always be the case (if a clustering parameter doesn't have a natural order or has a non-linear effect) or it might just be interesting to see a "data-driven" order of resolutions.
One way to come up with an ordering would be to calculate the Adjusted Rand Index between pairs of clustering. Potentially starting with the clustering with most clusters and greedily adding those with highest ARI to build up the tree.
Currently {clustree} assumes the user knows the order of resolutions but this might not always be the case (if a clustering parameter doesn't have a natural order or has a non-linear effect) or it might just be interesting to see a "data-driven" order of resolutions.
One way to come up with an ordering would be to calculate the Adjusted Rand Index between pairs of clustering. Potentially starting with the clustering with most clusters and greedily adding those with highest ARI to build up the tree.
Idea originally from Aaron Lun.