0

enter image description here

When viewing this, how do we know to find the optimal number of clusters? I used K-means and found the "elbow" on the graph that showed the optimal point but I am having trouble figuring this out from just the dendrogram.

rmahesh
  • 739
  • 2
  • 14
  • 30

1 Answers1

0

The interpretation varies depending on your metric and linkage used.

But in general, you want to keep branches that have "many" observations and with a “large" distance above (for the next merge).

Has QUIT--Anony-Mousse
  • 76,138
  • 12
  • 138
  • 194
  • So the ideal way to go would be to find the groupings that have the largest distance between each of the clusters? We use Euclidian distance as our metric. – rmahesh Nov 05 '17 at 21:52