index_label.append(label.index(i))
# print(index_label)
print(len(index_label),len(set(TAG_label)))
"""用多标记学习进行预测"""
k =[];k1=[]
k2=[];k3=[]
labelcount =1853
endianness = 'little'
feature_type = 'float'
encode_nominal = True
load_sparse = True
X, y1 = Dataset.load_arff_to_numpy("medical_data2.arff",
labelcount=labelcount,
endian="little",
input_feature_type=feature_type,
encode_nominal=encode_nominal,
load_sparse=load_sparse)
X=X.toarray()
#kmeans=KMeans(n_clusters=13,random_state=0).fit(X)
kmeans2=hcluster.fclusterdata(X, criterion='maxclust', t=12)
#kmeans = SpectralClustering(n_clusters=8).fit(X)
#print kmeans.labels_
index_0=[];index_1=[];index_2=[];index_3=[];index_4=[];index_5=[]
for i,li in enumerate(kmeans2):
if li==0:
index_0.append(i)
elif li==1:
index_1.append(i)
elif li==2:
# where the labels are located,
# big = at the beginning of the file
endianness = 'little'
# dtype used in the feature space
feature_type = 'float'
# whether the nominal attributes should be encoded as integers
encode_nominal = True
# if True - use the sparse loading mechanism from liac-arff
# if False - load dense representation and convert to sparse
load_sparse = True
# load data
X_train, y_train = Dataset.load_arff_to_numpy(
"path_to_data/dataset-train.dump.bz2",
labelcount=labelcount,
endian="big",
input_feature_type=feature_type,
encode_nominal=encode_nominal,
load_sparse=load_sparse)
X_test, y_test = Dataset.load_arff_to_numpy(
"path_to_data/dataset-train.dump.bz2",
labelcount=labelcount,
endian="big",
input_feature_type=feature_type,
encode_nominal=encode_nominal,
load_sparse=load_sparse)