You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
What's wrong?
For records that cannot be clustered, most clustering widgets produce a "?" in the Cluster column. Only Hierarchical Clustering assigns a cluster number to each record. However, when the dividing line is slid to the right, clusters of only one record may appear, which are in fact not clusters (or if they are, clusters of one item should also get their own cluster number from the other clustering widgets).
This is problematic when comparing different clustering approaches applied to the same dataset. Silhouette Plot, for instance gives a warning if there are "?"s in the Cluster column but treats one-item clusters from hierarchical clustering as true clusters.
I'd propose to let Hierarchical Clustering produce a "?" in the Cluster column if a "cluster" has only one record.
The text was updated successfully, but these errors were encountered:
What's wrong?
For records that cannot be clustered, most clustering widgets produce a "?" in the Cluster column. Only Hierarchical Clustering assigns a cluster number to each record. However, when the dividing line is slid to the right, clusters of only one record may appear, which are in fact not clusters (or if they are, clusters of one item should also get their own cluster number from the other clustering widgets).
This is problematic when comparing different clustering approaches applied to the same dataset. Silhouette Plot, for instance gives a warning if there are "?"s in the Cluster column but treats one-item clusters from hierarchical clustering as true clusters.
I'd propose to let Hierarchical Clustering produce a "?" in the Cluster column if a "cluster" has only one record.
The text was updated successfully, but these errors were encountered: