CODEWITHSUNDEEP
rcluster-analysishierarchical-clustering
Essi
Essi

Reputation: 801

R - Identify "Prime Examples" in Cluster Analysis

I have performed a Cluster Analysis and as a result I got, lets say 4 different clusters. With $kcmem I can see which Station belongs to which Cluster. But I want to pick out one Station out of each Cluster and analyse it further.

I got this result (its only a test, I have more than 6 Stations). And I picked out 4 Clusters.

$kcmem
     4 5
[1,] 1 1
[2,] 2 2
[3,] 3 3
[4,] 3 3
[5,] 4 4
[6,] 4 5

Of course I want to pick the Station that is the most typical one for each Cluster. How do I do that? From the main component analysis I know that actually there is a list of loadings and you can see how much a Station is related to a group.

I can only think that $kccentroids? But all Stations in the same Cluster have the same value then. 

$kccentroids
     [,1]      [,2]      [,3]      [,4]      [,5]      [,6]
4 1.0000000 0.5043308 0.7270394 0.7269328 0.6742015 0.6782120
4 0.5043308 1.0000000 0.5465914 0.5685687 0.6443715 0.5181649
4 0.7269861 0.5575801 0.9541015 0.9541015 0.8253911 0.8108125
4 0.6762068 0.5812682 0.7937889 0.8424147 0.9630949 0.9630949

Can anyone help? I would need a command that shows why a station belongs to a certain Cluster, so that I can pick out 1 Station of each Cluster as a "prime example".

Upvotes: 0

Views: 42

Answers (1)

Has QUIT--Anony-Mousse
Has QUIT--Anony-Mousse

Reputation: 77464

Choose the station closest to the cluster centroid.

That would be the usual approach for KMeans.

If you use algorithms such as PAM (k medoids) or affinity propagation or minimax then these methods will naturally provide a cluster representative.

Upvotes: 1

Related Questions