lmcinnes
commented
8 years ago There are a couple of tree structures. The first is the condensed tree. This has the clusters, with points falling out of them. You might have to do a little work to fit that with a heatmap, but it is doable. I would suggest you use the to_pandas method to get a dataframe and understand exactly what that data structure contains if you want to use it.
The second option is the single_linkage_tree which has a to_numpy method that will give you exactly the same dendrogram that scipy's single linkage clustering does (and is what seaborn uses internally in its clustermap). That might be easier to use immediately.
I reccommend the condensed tree, but as I say you might have to work through the details a bit to work out the best visualisation to go with a heatmap.
On Wed, Aug 17, 2016 at 7:50 PM, Tomasz notifications@github.com wrote:
I'm wondering if it is possible to get a tree-like structure of the resulting clustering. What I'm ultimately trying to do is to get a dendogram overlaid on top of a heat map (e.g. seaborn.clustermap)
From looking at the source code, it seems like there's a condensed tree attribute and that there's a _raw_tree attribute, is that what I should be looking at?
cc @ElDeveloper https://github.com/ElDeveloper
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/lmcinnes/hdbscan/issues/55, or mute the thread https://github.com/notifications/unsubscribe-auth/ALaKBWrMmdPp6P2NtKK9uMj313gXkSRDks5qg545gaJpZM4JnAC2 .
ElDeveloper
I'm wondering if it is possible to get a tree-like structure of the resulting clustering. What I'm ultimately trying to do is to get a dendogram overlaid on top of a heat map (e.g. seaborn.clustermap)
From looking at the source code, it seems like there's a condensed tree attribute and that there's a
_raw_treeattribute, is that what I should be looking at?cc @ElDeveloper