How can I use of Silhouette, Elbow, or other methods to specify the number of clusters?

[#1]

How can I use of Silhouette Method, Elbow Method, or other methods to specify the number of clusters by JMP?

[#2 · Accepted solution]

Hello @Asi,

Short answer: To my best knowledge, JMP doesn't use by default Silhouette Score or Elbow method. You'll have to write the script to use these methods.

Longer answer : In JMP, you have access to 3 clustering algorithms (+ latent class analysis and cluster variables):

- Hierarchical clustering,  

- K-means, where the criterion for best number of clusters k is CCC (Cubic Cluster Criterion), more infos here : https://documentation.sas.com/?docsetId=emref&docsetTarget=n1dm4owbc3ka5jn11yjkod7ov1va.htm&docsetVe... 

- Normal mixtures, where the criterion for best number of clusters is AICc (Corrected Akaike Information Criteron) and BIC (Bayesian Information Criterion).

You can find more infos on this excellent JMP Blog post : Clustering methods for unsupervised machine learning (jmp.com)
Hope this will help you !

[#3]

Hello Victor, Thanks a lot for the good solution!