2

I have two distance matrices: one produced based on cosine distances and another produced with Gower distance. I have used these separately for clustering, but I would like to combine them somehow to feed into a single clustering model.

I have read about two options to combine distance matrices: do a weighted sum (w*D1+(1-w)*D2) or superimpose (like a 3D object) and average them point by point. I am inclined towards the first, but I don't know if these 2 distances have the correct properties for the weighted sum.

Can you clarify what is the common practice? Is there any R package to handle such cases?

Sapiens
  • 224
  • First of all, define please "cosine distance" and "Gower distance", for both coefficients are similarities (Read of some "gower distances" e.g. here https://stats.stackexchange.com/a/15313/3277). – ttnphns Nov 09 '23 at 20:02
  • Second, what are your data types (explain, describe) and what makes you to desire to blend the different proximity measures? – ttnphns Nov 09 '23 at 20:04

1 Answers1

0

The HclustCompro method might help you with that. It is implemented in the SPARTAAS R package. Here is an article using an adaptation of the method: https://onlinelibrary.wiley.com/doi/full/10.1002/sim.9625

klg
  • 1
  • While this link may answer the question, it is better to include the essential parts of the answer here and provide the link for reference. Link-only answers can become invalid if the linked page changes. - From Review – Shawn Hemelstrand Jan 26 '24 at 11:48