def main_subspace_kmeans():
""" Subspace K-means
"""
from sklearn.preprocessing import MinMaxScaler
nrOfClusters = 4
handler = SubKmeans()
result = handler.runWithReplicates(data, nrOfClusters, 10)
label_ = result.labels
from munkres import Munkres
def maplabels(L1, L2):
Label1 = np.unique(L1)
Label2 = np.unique(L2)
nClass1 = len(Label1)
nClass2 = len(Label2)
nClass = np.maximum(nClass1, nClass2)
print("nClass:", nClass)
G = np.zeros((nClass, nClass))
for i in range(nClass1):
ind_cla1 = L1 == Label1[i]
ind_cla1 = ind_cla1.astype(float)
for j in range(nClass2):
ind_cla2 = L2 == Label2[j]
ind_cla2 = ind_cla2.astype(float)
G[i, j] = np.sum(ind_cla2 * ind_cla1)
m = Munkres()
index = m.compute(-G.T)
index = np.array(index) # print(-G.T)
print("map rule:\n", index)
temp = np.array(L2)
newL2 = np.zeros(temp.shape, dtype=int)
for i in range(nClass2):
for j in range(len(temp)):
if temp[j] == index[i, 0]:
newL2[j] = index[i, 1]
return newL2
label_ = maplabels(label, label_)
print("accuracy")
from sklearn.metrics import accuracy_score
print(accuracy_score(label, label_))
print("nmi")
from sklearn.metrics import normalized_mutual_info_score
print(normalized_mutual_info_score(label, label_))
print("Rand")
from sklearn.metrics import adjusted_rand_score
print(adjusted_rand_score(label, label_))
'''
nmi = NormalizedMutualInformation.forLabelSeq(label, label_)
print(f"NMI: {nmi}")
print(f"m: {result.m()}")
'''
transformedData = np.matmul(result.transformation().T,
data[:, :, None]).squeeze()
DataIO.writeClusters(
Path("./result") / f"{dataname}_result.dat", transformedData,
result.labels)
def main(ds, folder, separator=";"):
dataset = Path(f"./{folder}/{ds}.dat")
dpSeq, labelSeq = DataIO.loadCsvWithIntLabelsAsSeq(dataset, separator=separator)
df = pd.DataFrame(dpSeq[:, :5])
cl = pd.Series(labelSeq)
factor_scatter_matrix(df, cl)
plt.show()
def main_subspace_kmeans(ds, separator=";"):
""" Subspace K-means
"""
dataset = Path(f"./datasets/{ds}.dat")
dataname = dataset.stem
d, groundTruth = DataIO.loadCsvWithIntLabelsAsSeq(dataset,
separator=separator)
data = DataNormalization.standardizeData(d)
nrOfClusters = np.unique(groundTruth).shape[0]
handler = SubKmeans()
result = handler.runWithReplicates(data, nrOfClusters, 10)
nmi = NormalizedMutualInformation.forLabelSeq(groundTruth, result.labels)
print(f"NMI: {nmi}")
print(f"m: {result.m()}")
transformedData = np.matmul(result.transformation().T,
data[:, :, None]).squeeze()
DataIO.writeClusters(
Path("./result") / f"{dataname}_result.dat", transformedData,
result.labels)
def main_kmeans(ds, separator=";"):
""" K-means
"""
dataset = Path(f"./datasets/{ds}.dat")
d, groundTruth = DataIO.loadCsvWithIntLabelsAsSeq(dataset,
separator=separator)
data = DataNormalization.standardizeData(d)
nrOfClusters = np.unique(groundTruth).shape[0]
all_nmi = 0
for _ in range(10):
result_labels = k_means(data, nrOfClusters)
nmi = NormalizedMutualInformation.forLabelSeq(groundTruth,
result_labels)
all_nmi += nmi
print(f"NMI: {all_nmi / 10}")
def main_subspace_kmeans(ds, separator=";"):
""" Subspace K-means
"""
dataset = Path(f"./datasets/{ds}.dat")
dataname = dataset.stem
from sklearn.preprocessing import LabelEncoder
df[df.columns[-1]] = LabelEncoder().fit_transform(df[df.columns[-1]])
data = df[df.columns[0:-1]]
label = df[df.columns[-1]]
from sklearn.preprocessing import MinMaxScaler
data = MinMaxScaler().fit_transform(data)
#d, groundTruth = DataIO.loadCsvWithIntLabelsAsSeq(dataset, separator=separator)
#data = DataNormalization.standardizeData(d)
nrOfClusters = len(label.unique())
# nrOfClusters = np.unique(groundTruth).shape[0]
#print("类别：",nrOfClusters)
handler = SubKmeans()
result = handler.runWithReplicates(data, nrOfClusters, 10)
label_ = result.labels
from munkres import Munkres
def maplabels(L1, L2):
Label1 = np.unique(L1)
#print("Label1:",Label1)
Label2 = np.unique(L2)
#print("Label2:",Label2)
nClass1 = len(Label1)
nClass2 = len(Label2)
#print("nClass1:",nClass1)
#print("nClass2:",nClass2)
nClass = np.maximum(nClass1, nClass2)
#print("nClass:",nClass)
G = np.zeros((nClass, nClass))
for i in range(nClass1):
ind_cla1 = L1 == Label1[i]
ind_cla1 = ind_cla1.astype(float)
for j in range(nClass2):
ind_cla2 = L2 == Label2[j]
ind_cla2 = ind_cla2.astype(float)
G[i, j] = np.sum(ind_cla2 * ind_cla1)
m = Munkres()
index = m.compute(-G.T)
index = np.array(index) # print(-G.T)
#print("map rule:\n", index)
temp = np.array(L2)
newL2 = np.zeros(temp.shape, dtype=int)
for i in range(nClass2):
for j in range(len(temp)):
if temp[j] == index[i, 0]:
newL2[j] = index[i, 1]
return newL2
#print("result_acc:",result_acc)
# print(type(groundTruth))
#print("groundTruth",groundTruth)
#print("label:",label)
#print("label_",label_)
label_ = maplabels(label, label_)
#print("newlabel:",label)
# print("newlabel_",label_)
print("accuracy")
from sklearn.metrics import accuracy_score
print(accuracy_score(label, label_))
print("nmi")
from sklearn.metrics import normalized_mutual_info_score
print(normalized_mutual_info_score(label, label_))
print("Rand")
from sklearn.metrics import adjusted_rand_score
print(adjusted_rand_score(label, label_))
'''
nmi = NormalizedMutualInformation.forLabelSeq(label, label_)
print(f"NMI: {nmi}")
print(f"m: {result.m()}")
'''
transformedData = np.matmul(result.transformation().T,
data[:, :, None]).squeeze()
DataIO.writeClusters(
Path("./result") / f"{dataname}_result.dat", transformedData,
result.labels)