0

There are several threads discussing clustering analysis of a distance matrix and they dismiss use of the k-means algorithm. Here are two examples:

However, it is hard to come by an intuitive (and ideally visual) explanation why it is wrong to apply k-means to a distance matrix. It will be really helpful is someone can explain this.

Krrr
  • 476
  • 6
  • 15
  • 1
    Your first link makes the basic point that you cannot calculate a *mean* without values to average. So it suggested $k$-*mediods* instead – Henry Jun 10 '21 at 17:18
  • My answer to the question in the first link says that it is possible (and how) to apply k-means to a distance matrix and that it is implied that the distances are euclidean. – ttnphns Jun 13 '21 at 10:13

0 Answers0