Hi everyone, I know that this topic was already mentioned in several post:
Evaluation of Clustering results
Alternative Methods for Cluster Validity
Validate the NbClust package use in Knime
And I consider that is important to have Knime native nodes to validate clustering results and this knowing before hand that it is possible to use R or Python to make such evaluations using the integration provided by the software and existence of the X-means Weka node, but counting with such validation indices (i) you can determine the “proper” number of clusters with a decent amount of data (as sometimes I get stuck), (ii) you can guarantee that the number of clusters that are suggested by the indices are the correct ones and not using the suggested number by R and Python and then apply that number in the Knime (i.e k-means node ) with the risk that you are not using the same configuration in both platforms such as diferent distances or something else and (iii) finally you can find the “best” cluster that fit your data using different algorithms and distances with Knime native cluster nodes ( at least with k-means and hierarchical clustering, dunno with Optics and Dbscan).
For that reason I beg you to include such validation measures, there are at least 30 indices acording to the NbClust package in R.
Thank you
Mau