Comparing kmeans() and hclust()
Comparing k-means and hierarchical clustering, you'll see the two methods produce different cluster memberships. This is because the two algorithms make different assumptions about how the data is generated. In a more advanced course, we could choose to use one model over another based on the quality of the models' assumptions, but for now, it's enough to observe that they are different.
This exercise will have you compare results from the two models on the pokemon
dataset to see how they differ.
This exercise is part of the course
Unsupervised Learning in R
Exercise instructions
The results from running k-means clustering on the pokemon
data (for 3 clusters) are stored as km.pokemon
. The hierarchical clustering model you created in the previous exercise is still available as hclust.pokemon
.
- Using
cutree()
onhclust.pokemon
, assign cluster membership to each observation. Assume three clusters and assign the result to a vector calledcut.pokemon
. - Using
table()
, compare cluster membership between the two clustering methods. Recall that the different components of k-means model objects can be accessed with the$
operator.
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Apply cutree() to hclust.pokemon: cut.pokemon
# Compare methods
table(___, ___)