Altair slc k mediods clustering using r
Too long to post on a list,see github
https://github.com/rogerjdeangelis/utl-altair-slc-k-mediods-clustering-using-r
Graphic output
The graph uses factor analysis to reduce the dimensionality for visualization of four dimesional space.
https://github.com/rogerjdeangelis/utl-altair-slc-k-mediods-clustering-using-r/blob/main/mediods.png
https://github.com/rogerjdeangelis/utl-altair-slc-k-mediods-clustering-using-r/blob/main/silhouette.png
PROBLEM IDENTIFY THE FOUR DIEMSIONAL MEDOIDS FOR THREE CLUSTERS
ONE APPLICATION: PROFILE PATIENTS INTO GROUPS
For the iris data the medoids are the observered points in four dimensional space,
Sepal_Length, Sepal_Width, Petal_Length, and Petal_Width,
that minimize the distance to all other points in the cluster.
HERE ARE THE FOUR DIMENSIONAL OBSERVED MEDIODS
SEPAL_ SEPAL_ PETAL_ PETAL_
CLUSTER LENGTH WIDTH LENGTH WIDTH
1 5.0 3.4 1.5 0.2 Probably setosa
2 6.0 2.9 4.5 1.5 Probably versicolor
3 6.8 3.0 5.5 2.1 Probably virginica
Note we are tring to identify three cluster without using the species variable.
We are using medoids based clusters, handles outliers better?
Species is hardcoded for display purposes only, species was not used to identify clusters.
I HARDCORED THE SPECIES, COULD HAVE LABELED CATEGORY1-3.
CLUSTER SETOSA VERSICOLOR VIRGINICA
1 50 0 0
2 0 48 14
3 0 2 36
community post
https://community.altair.com/discussion/49353/k-medoids-in-rapidminer/p1?tab=all
