Cluster algorithm for n-dimensional data and a given number of clusters

clusterize.rb

# A
#   B
#
#               D
#
#         C
#             E
#
# F
#
# # # # # # # # # # #
# 1 2 3 4 5 6 7 8 9 10

#http://www.psychstat.missouristate.edu/multibook/mlt04.htm

data = { [1,1] => [:A], [2,2] => [:B], [5,6] => [:C],
         [8,4] => [:D], [7,7] => [:E], [1,9] => [:F] }

# Euclidean distance between two n-dimensional points
def distance(a,b)
  Math.sqrt(a.zip(b).reduce(0.0){|s,e| s+=(e[0]-e[1])**2})
end

# Center point between two n-dimensional points
def centroid(a,b)
  a.zip(b).map!{|e| ((e[0]+e[1]) / 2.0).round }
end

# Iteratively create clusters inside n-dimendional data
def clusterize!(h)
  a,b = h.keys.combination(2).min_by{|e| distance(*e)}
  v = h[a] + h[b]
  h.delete a
  h.delete b
  h[centroid(a,b)] = v
end

p data
# {[1, 1]=>[:A], [2, 2]=>[:B], [5, 6]=>[:C], [8, 4]=>[:D], [7, 7]=>[:E], [1, 9]=>[:F]}
# let's create 3 clusters ...
clusterize!(data) while data.size > 3
p data
# {[1, 9]=>[:F], [2, 2]=>[:A, :B], [7, 6]=>[:D, :C, :E]}