Dear all,
I just wanted to share a couple of new k-Means components I have just added to the Hub.
The first one is k-Means (distance): it’s an extension of the native k-Means node, which outputs the euclidean distance between every node and its centroid, or the centroid of each cluster.
The second one is Same-size k-Means: it uses distances to force all clusters to have an equal number of points. Of course, by adding the “same-size” constraint, the clustering quality might degrade and be particularly prone to outliers. Still, in some circumstances, the component might be handy.
I’ve noticed that some older posts in the forum required these features, although they are closed now, so I can’t respond to them:
- K-means with equal cluster size
- k-means clustering: distance from a certain datapoint to his next centroid
Any inputs and feedback are welcome!
Andrea