K selection in clustering

[khanetor]

In clustering algorithms such as Kmeans or Kshape, is there a function to calculate the silhouette score to determine the best k value?

[fkiraly]

As far as I know, this does not exist in sktime, although it is very close to existing patterns for classification.

The sktime clusterers are compatible with sklearn if numpy based data containers are used - so if you can find the above implemented for sklearn clusterers, the naive combination with sktime clusters should, hopefully, work out of the box.

Would appreciate pointers to:

• implementations of the above for sklearn clusterers
• if you find these try, whether this works out of the box in combination for sktime