NO
di.classLabel = DNum
end
if error(Mi) >0.5
for i = i to k
d1 = 1/d
if all clussterNumber exist
Export model using training set
for j = 1 to d
update value
di.classLabel = clusterNumber
clusterNumber = kmeans(di)
calculation error
start
Standard weight
YES
Sampling with return
if di value is missing
for each di
