def buildCluster(self):
#Initialize each content as a Hierachical Cluster
hiclusters = [ HierachicalCluster(vec = self.__content[i], id = i ) for i in range(len(self.__content)) ]
distances = {}
flag = None;
currentclusted = -1
while(len(hiclusters) > 1) :
min_val = 2;
hiclusters_len = len(hiclusters)
'''
calculate distances between each two clusters,
if d<minval,record the two clusters' id
'''
for i in range(hiclusters_len-1) :
for j in range(i + 1, hiclusters_len) :
if distances.get((hiclusters[i].id,hiclusters[j].id)) == None:
distances[(hiclusters[i].id,hiclusters[j].id)] = ClusterBase.pearson_distance(hiclusters[i].vec,hiclusters[j].vec)
d = distances[(hiclusters[i].id,hiclusters[j].id)]
if d < min_val :
min_val = d
flag = (i,j)
bic1,bic2 = flag
newvec = [(hiclusters[bic1].vec[i] + hiclusters[bic2].vec[i])/2 for i in range(len(hiclusters[bic1].vec))]
newbic = HierachicalCluster(newvec, left=hiclusters[bic1], right=hiclusters[bic2], distance=min_val, id = currentclusted)
currentclusted -= 1
del hiclusters[bic2]
del hiclusters[bic1]
hiclusters.append(newbic)
return hiclusters[0]
def buildCluster(self):
ClusterBase.buildCluster(self)
min_max_per_word = [ [min([row[i] for row in self.__content]), max([row[i] for row in self.__content])] for i in range(len(self.__content[0]))]
# generate k clusters randomly
__ClusterCenters = []
for i in range(self.__KCluster) :
cluster = []
for min_, max_ in min_max_per_word :
cluster.append(random.random() * (max_ - min_) + min_)
__ClusterCenters.append(cluster)
lables = []
matchs = [ [] for i in range(self.__KCluster)]
lastmatchs = [ [] for i in range(self.__KCluster)]
rounds = 100
while rounds > 0 :
matchs = [ [] for i in range(self.__KCluster)]
print('round \t', rounds)
for i in range(len(self.__content)) :
bestmatch_cluster = None
min_distance = 2.1
for j in range(self.__KCluster) :
dis = ClusterBase.pearson_distance(__ClusterCenters[j], self.__content[i])
if dis < min_distance :
min_distance = dis
bestmatch_cluster = j
matchs[bestmatch_cluster].append(i)
self.printCluster(matchs)
self.printCluster(lastmatchs)
if matchs == lastmatchs : break
lastmatchs = [[ item for item in matchs[i] ] for i in range(self.__KCluster)]
# move the centroids to the average of their members
for j in range(self.__KCluster) :
avg = [0.0 for i in range(len(self.__content[0])) ]
for m in matchs[j] :
vec = self.__content[m]
for i in range(len(self.__content[0])) :
avg[i] += vec[i]
avg = [ item / len(self.__content[0]) for item in avg]
__ClusterCenters[j] = avg
rounds -= 1
