I was considering using the silhouette score, however its complexity is $O(mn^2)$ which can quickly become prohibitive with the size of the dataset.
$\begingroup$
$\endgroup$
2
-
$\begingroup$ Are you asking about some internal clustering validation criteria? If yes, see stats.stackexchange.com/a/358937/3277. Pay attention to Clalinski-Harabasz, Davies-Bouldin. $\endgroup$– ttnphnsCommented Sep 20, 2018 at 11:31
-
$\begingroup$ Thank you. Are there references listing the complexity of those different validation metrics (my computational ressources are limited...)? $\endgroup$– TanguyCommented Sep 20, 2018 at 18:29
Add a comment
|