def clustering(self, K, N=100):
classes = {}
# generate K initial points
pts = []
for i in range(K):
classes[i] = []
pt = []
for j in range(len(self.data[0][0])):
pt.append(random.random()) # normallized data
pts.append(pt)
#
for n in range(N):
for i in range(K):
classes[i] = []
for i in range(len(self.data)):
temp = 9999
flag = 0
for j in range(K):
dist = Distance.euclideanDist(self.data[i][0], pts[j])
if dist < temp:
temp = dist
flag = j
classes[flag].append(i)
#
for i in range(K):
means = [0]*len(self.data[0][0])
for j in range(len(classes[i])):
for k in range(len(self.data[0][0])):
means[k] = means[k]+self.data[j][0][k]
for k in range(len(self.data[0][0])):
means[k] = means[k]/len(self.data[0][0])
pts[i] = means
return classes
def predict(self, sample, k):
candidats = [{9999: 0}] * k
ids = [0] * k
for i in range(len(self.tdata)):
dist = Distance.euclideanDist(sample, self.tdata[i][0])
if dist < candidats[-1].keys()[0]:
candidats.sort()
candidats.pop()
candidats.insert(0, {dist: i})
# most is the winner
counter = {}
for c in candidats:
counter[self.tdata[c.values()[0]][1]] = 0
for c in candidats:
counter[self.tdata[c.values()[0]][1]] = counter[self.tdata[c.values()[0]][1]] + 1
print counter
