def rsnn(sampledData, remainedData, sampledIndex, remainedIndex, singleName):
predicted_labelAll = []
for i in range(len(sampledData)):
# clusters = random.randint(min_clusters,max_clusters)
clusters = random.randint(2, 11)
# clusters = random.randint(2,11)#范围是[2,10]
if singleName == 'kmeans':
predicted_label = KMeans(n_clusters=clusters).fit_predict(
sampledData[i])
elif singleName in ('ward', 'complete', 'average'):
predicted_label = AgglomerativeClustering(
linkage=singleName,
n_clusters=clusters).fit_predict(sampledData[i])
predicted_labelAll.append(predicted_label.tolist()) ##对采样出来的数据集的预测标签集合
assinALLNnLabels = [] #全部的通过近邻分配的标签
#remainedData和sampleedData拥有的数据的行数是一致的，所以j的值无论从len(remainedData)还是从len(sampledData)取都可以
for j in range(len(remainedData)):
assinNnLabels = [] # 通过近邻分配的标签
for m in range(len(remainedData[j])):
minDist = inf
minindex = -1
for k in range(len(sampledData[j])):
distJI = distEclud(remainedData[j][m], sampledData[j][k])
if distJI < minDist:
minDist = distJI
minindex = k
assinNnLabels.append(
predicted_labelAll[j][minindex]) #对除采样外的数据集的根据近邻关系得到的预测标签集合
assinALLNnLabels.append(assinNnLabels)
#对两个预测标签和序列值分别进行组合
combineIndex = []
combinedLables = []
for column in range(len(predicted_labelAll)):
combineIndexOne = sampledIndex[column] + remainedIndex[column]
combinedLablesOne = predicted_labelAll[column] + assinALLNnLabels[
column]
combineIndex.append(combineIndexOne)
combinedLables.append(combinedLablesOne)
#把打乱的序号按照从小到大排列出来，得到元素升序的序列值
seqIndexAll = []
for combineIndex1 in combineIndex:
seqIndex = []
for seq in range(len(sampledData[0]) + len(remainedData[0])):
for elementIndex in range(len(combineIndex1)):
if combineIndex1[elementIndex] == seq:
seqIndex.append(elementIndex)
seqIndexAll.append(seqIndex)
#得到真正的sampledData和remainedData组合后的标签值
finalLabel = []
for finalIndex in range(len(combinedLables)):
finallabelone = []
for index in seqIndexAll[finalIndex]:
finallabelone.append(combinedLables[finalIndex][index])
finalLabel.append(finallabelone) #最终聚类结果
return finalLabel
def fsrsnn(sampledData, remainedData, sampledIndex, remainedIndex,
sampledDataFs, k):
min_clusters, max_clusters = k_range(k) # 根据真实类标签数得到实验所用的簇数量范围
predicted_labelAll = []
for i in range(len(sampledData)):
clusters = random.randint(min_clusters, max_clusters)
# clusters = random.randint(2,11)#范围是[2,10]
predicted_label = KMeans(n_clusters=clusters).fit_predict(
sampledDataFs[i])
predicted_labelAll.append(predicted_label.tolist()) ##对采样出来的数据集的预测标签集合
assinALLNnLabels = [] #全部的通过近邻分配的标签
#remainedData和sampleedData拥有的数据的行数是一致的，所以j的值无论从len(remainedData)还是从len(sampledData)取都可以
for j in range(len(remainedData)):
assinNnLabels = [] # 通过近邻分配的标签
for m in range(len(remainedData[j])):
minDist = inf
minindex = -1
for k in range(len(sampledData[j])):
distJI = distEclud(remainedData[j][m],
sampledData[j][k]) # 计算质心和数据点之间的距离
if distJI < minDist:
minDist = distJI
minindex = k
assinNnLabels.append(
predicted_labelAll[j][minindex]) #对除采样外的数据集的根据近邻关系得到的预测标签集合
assinALLNnLabels.append(assinNnLabels)
#对两个预测标签和序列值分别进行组合
combineIndex = []
combinedLables = []
for column in range(len(predicted_labelAll)):
combineIndexOne = sampledIndex[column] + remainedIndex[column]
combinedLablesOne = predicted_labelAll[column] + assinALLNnLabels[
column]
combineIndex.append(combineIndexOne)
combinedLables.append(combinedLablesOne)
#把打乱的序号按照从小到大排列出来，得到元素升序的序列值
seqIndexAll = []
for combineIndex1 in combineIndex:
seqIndex = []
for seq in range(len(sampledData[0]) + len(remainedData[0])):
for elementIndex in range(len(combineIndex1)):
if combineIndex1[elementIndex] == seq:
seqIndex.append(elementIndex)
seqIndexAll.append(seqIndex)
#得到真正的sampledData和remainedData组合后的标签值
finalLabel = []
for finalIndex in range(len(combinedLables)):
finallabelone = []
for index in seqIndexAll[finalIndex]:
finallabelone.append(combinedLables[finalIndex][index])
finalLabel.append(finallabelone) #最终聚类结果
return finalLabel
def initialMultiRun(data, times, singleName):
predicted_labelAll = []
for i in range(times):
clusters = random.randint(2, 11)
if singleName == "kmeans":
predicted_label = KMeans(n_clusters=clusters).fit_predict(data)
elif singleName in ('ward', 'average', 'complete'):
predicted_label = AgglomerativeClustering(
linkage=singleName, n_clusters=clusters).fit_predict(data)
predicted_labelAll.append(predicted_label.tolist())
return predicted_labelAll
def main():
predicted_labelAll = []
datamat, datalabels = loadDataset("../dataset/iris.data")
print 'data ready'
nmi_max = -inf
ari_max = -inf
for i in range(10):
clusters = random.randint(2, 11)
predicted_label = KMeans(n_clusters=clusters).fit_predict(datamat)
predicted_label = predicted_label.tolist()
nmi = normalized_mutual_info_score(datalabels, predicted_label)
ari = adjusted_rand_score(datalabels, predicted_label)
if nmi > nmi_max:
nmi_max = nmi
if ari > ari_max:
ari_max = ari
print('nmi值为：')
print(nmi_max)
print('ari值为：')
print(ari_max)
