1

I have agglomerative clustering done.

I want to convert it to a decision tree so I can figure out the cluster very quickly.

How to do so?

A tedious approach (bad, I know):

  • Take the top agglomeration.
  • Train a classifier for the agglomeration.
  • Iterate this for all agglomerations.
Karel Macek
  • 2,463
  • 11
  • 23
  • 1
    Generally, to train agglomerative clustering dendrogram by a supervised method is not a good idea. Most of the time and linkage methods, agglomeration is just a history of collection and not a clear-cut hierarchical classification. Because (i, practically) on some levels the decision to merge this pair of clusters is just minorily better than to merge some other pair: so you won't succeed to train to do it by a classifier. – ttnphns Nov 14 '18 at 11:13
  • ...(ii, theoretically), as already said, dendrogram may be not a classification with emergent cluster properties, philosophically (see point [here](https://stats.stackexchange.com/a/63549/3277)). – ttnphns Nov 14 '18 at 11:13

0 Answers0