def update_clustroid(cluster):
"""Update the clustroid in a cluster using the minimum sum of squares"""
sum_of_squares = []
for i in xrange(len(cluster)):
summ = 0
for j in xrange(i + 1, len(cluster)):
summ += pow(edit_dist(cluster[i], cluster[j]), 2)
sum_of_squares.append(summ)
c_id = argmin(sum_of_squares)
cluster.append(cluster[c_id])
del cluster[c_id]
def set_clustroid_mate(listoclusters):
"""For each cluster, set the [-2] element as the closest to the clustroid
"""
min_dist = float('inf')
for cluster in listoclusters:
if len(cluster) == 1:
continue
for i in xrange(len(cluster) - 1):
dist = edit_dist(cluster[i], cluster[-1])
if min_dist > dist:
min_dist = dist
mate_idx = i
cluster[-2], cluster[mate_idx] = cluster[mate_idx], cluster[-2]
def clusterize(listoclusters):
"""Create clusters starting from a list of lists of single elements"""
cluster_size = 1
while cluster_size < settings.CLUSTER_MAX_SIZE:
min_dist = float('inf')
for i in xrange(len(listoclusters)):
for j in xrange(i + 1, len(listoclusters)):
dist = edit_dist(listoclusters[i][-1], listoclusters[j][-1])
if dist < min_dist:
min_dist = dist
to_merge = (i, j)
listoclusters[to_merge[0]].extend(listoclusters[to_merge[1]])
update_clustroid(listoclusters[to_merge[0]])
cluster_size = max(len(l) for l in listoclusters)
set_clustroid_mate(listoclusters)
