# As the initial centroids are defined randomly,
# we define a seed for purposes of reprodutability
set.seed(123)

# Let's remove the column with the mammals' names, so it won't be used in the clustering
input <- all.mammals.milk.1956[,2:6]

# The nstart parameter indicates that we want the algorithm to be executed 20 times.
# This number is not the number of iterations, it is like calling the function 20 times and then
# the execution with lower variance within the groups will be selected as the final result.
kmeans(input, centers = 3, nstart = 20)