def test_if_works_with_cross_validation(self):
# scikit does not support sparse output matrix in CV
classifier = RakelO(base_classifier=GaussianNB(),
base_classifier_require_dense=[True, True],
labelset_size=TEST_LABELSET_SIZE,
model_count=TEST_MODEL_COUNT,
require_dense=[False, True])
self.assertClassifierWorksWithCV(classifier)
def RAkELO(dataset_train_x, dataset_train_y, dataset_test_x, dataset_test_y,
base_clasif, num_labels, num_models):
classifier = RakelO(base_classifier=base_clasif,
labelset_size=num_labels,
model_count=num_models)
classifier.fit(dataset_train_x, dataset_train_y)
predictions = classifier.predict(dataset_test_x)
Metrics_Accuracy("RAkELO", predictions, dataset_test_y)
def RAkELO(dataset_train_x, dataset_train_y, dataset_test_x, dataset_test_y,
base_clasif, num_labels, num_models):
classifier = RakelO(base_classifier=base_clasif,
labelset_size=num_labels,
model_count=num_models)
start_time = time.time()
classifier.fit(dataset_train_x, dataset_train_y)
stop_time = time.time()
time_lapsed = stop_time - start_time
predictions = classifier.predict(dataset_test_x)
Metrics_Accuracy("RAkELO", predictions, dataset_test_y)
print("Execution time: {}s".format(time_lapsed))
def getClassifier(self):
if self.classifierType.lower() == 'rakelo':
classifier = RakelO(
base_classifier=LabelPowerset(GaussianNB()),
#base_classifier_require_dense=[True, True],
model_count=10,
labelset_size=2 #len(labelTypes) // 4
)
elif self.classifierType.lower() == 'mlknn':
classifier = MLkNN(k=3)
# elif self.classifierType.lower() == 'mltsvm':
# classifier = MLTSVM(c_k = 2**-1)
elif self.classifierType.lower() == 'mlaram':
classifier = MLARAM()
elif self.classifierType.lower() == 'labelpowerset':
classifier = LabelPowerset(
classifier=RandomForestClassifier(n_estimators=100),
require_dense=[False, True])
return classifier
def get_rakeld_with_nb(self):
return RakelO(base_classifier=GaussianNB(),
base_classifier_require_dense=[True, True],
labelset_size=TEST_LABELSET_SIZE,
model_count=TEST_MODEL_COUNT)
def get_rakeld_with_svc(self):
return RakelO(base_classifier=SVC(),
base_classifier_require_dense=[False, True],
labelset_size=TEST_LABELSET_SIZE,
model_count=TEST_MODEL_COUNT)
def Util_ClassifierMethods(dataset_train_x, dataset_train_y, dataset_test_x,
dataset_test_y):
#BR
Util_Title("Binary Relevance")
base_classif = GaussianNB()
BinaryRelevance(dataset_train_x, dataset_train_y, dataset_test_x,
dataset_test_y, base_classif, "GaussianNB")
dict_res = FindBestSVCParams(skpt.BinaryRelevance(), dataset_train_x,
dataset_train_y)
base_classif = SVC(kernel=dict_res['classifier__kernel'],
degree=dict_res['classifier__degree'])
BinaryRelevance(dataset_train_x, dataset_train_y, dataset_test_x,
dataset_test_y, base_classif, "SVC tuned")
dict_res = FindBestMNBParams(skpt.BinaryRelevance(), dataset_train_x,
dataset_train_y)
base_classif = MultinomialNB(alpha=dict_res['classifier__alpha'])
BinaryRelevance(dataset_train_x, dataset_train_y, dataset_test_x,
dataset_test_y, base_classif, "MNB tuned")
#CC
Util_Title("Classifier Chain")
base_classif = GaussianNB()
ClassifierChain(dataset_train_x, dataset_train_y, dataset_test_x,
dataset_test_y, base_classif, "GaussianNB")
dict_res = FindBestSVCParams(skpt.ClassifierChain(), dataset_train_x,
dataset_train_y)
base_classif = SVC(kernel=dict_res['classifier__kernel'],
degree=dict_res['classifier__degree'])
ClassifierChain(dataset_train_x, dataset_train_y, dataset_test_x,
dataset_test_y, base_classif, "SVC tuned")
dict_res = FindBestMNBParams(skpt.ClassifierChain(), dataset_train_x,
dataset_train_y)
base_classif = MultinomialNB(alpha=dict_res['classifier__alpha'])
ClassifierChain(dataset_train_x, dataset_train_y, dataset_test_x,
dataset_test_y, base_classif, "MNB tuned")
#LP
Util_Title("Label Powerset")
base_classif = GaussianNB()
LabelPowerset(dataset_train_x, dataset_train_y, dataset_test_x,
dataset_test_y, base_classif, "GaussianNB")
dict_res = FindBestSVCParams(skpt.LabelPowerset(), dataset_train_x,
dataset_train_y)
base_classif = SVC(kernel=dict_res['classifier__kernel'],
degree=dict_res['classifier__degree'])
LabelPowerset(dataset_train_x, dataset_train_y, dataset_test_x,
dataset_test_y, base_classif, "SVC tuned")
dict_res = FindBestMNBParams(skpt.LabelPowerset(), dataset_train_x,
dataset_train_y)
base_classif = MultinomialNB(alpha=dict_res['classifier__alpha'])
LabelPowerset(dataset_train_x, dataset_train_y, dataset_test_x,
dataset_test_y, base_classif, "MNB tuned")
#MLkNN
Util_Title("MLkNN")
dict_res = FindBestK(skadapt.MLkNN(), dataset_train_x, dataset_train_y)
MLkNN(dataset_train_x, dataset_train_y, dataset_test_x, dataset_test_y,
dict_res['k'], dict_res['s'])
#MLARAM
Util_Title("MLARAM")
dict_res = FindBestVT(dataset_train_x, dataset_train_y)
MLARAM(dataset_train_x, dataset_train_y, dataset_test_x, dataset_test_y,
dict_res['vigilance'], dict_res['threshold'])
#BRkNNa
Util_Title("BRkNNa")
dict_res = FindBestK(skadapt.BRkNNaClassifier(), dataset_train_x,
dataset_train_y)
BRkNNa(dataset_train_x, dataset_train_y, dataset_test_x, dataset_test_y,
dict_res['k'])
#BRkNNb
Util_Title("BRkNNb")
dict_res = FindBestK(skadapt.BRkNNbClassifier(), dataset_train_x,
dataset_train_y)
BRkNNb(dataset_train_x, dataset_train_y, dataset_test_x, dataset_test_y,
dict_res['k'])
#RAkELD
Util_Title("RAkELd")
dict_res = GridSearchCV_baseRakel(RakelD(), dataset_train_x,
dataset_train_y)
RAkELd(dataset_train_x, dataset_train_y, dataset_test_x, dataset_test_y,
dict_res['base_classifier'], dict_res['labelset_size'])
#RAkELo
Util_Title("RAkELo")
dict_res = GridSearchCV_baseRakel(RakelO(), dataset_train_x,
dataset_train_y)
RAkELO(dataset_train_x, dataset_train_y, dataset_test_x, dataset_test_y,
dict_res['base_classifier'], dict_res['labelset_size'],
dict_res['model_count'])
#MLTSVM
Util_Title("MLTSVM")
dict_res = FindCKParam(dataset_train_x, dataset_train_y, dataset_test_x,
dataset_test_y)
TwinMLSVM(dataset_train_x, dataset_train_y, dataset_test_x, dataset_test_y,
dict_res['c_k'], dict_res['sor_omega'])
def GridSearchCV_baseRakel(classif, dataset_train_x, dataset_train_y):
#labelset_size denotes the desired size of partition
range_labelset_size = list(range(1, 11))
rangefloat = [round(x * 0.1, 1) for x in range(1, 11)]
parameters = [
{
'base_classifier': [GaussianNB()],
'labelset_size': range_labelset_size,
},
{
'base_classifier': [MultinomialNB()],
'base_classifier__alpha':
rangefloat, #for smoothing {Additive smoothing parameter NB}
'labelset_size': range_labelset_size,
},
{
'base_classifier': [SVC()],
'base_classifier__kernel': ['rbf', 'linear', 'sigmoid'],
'labelset_size': range_labelset_size,
},
]
print(type(classif) == type(RakelO()))
if (type(classif) == type(RakelO())):
end_range = dataset_train_y.shape[1] // 2 if dataset_train_y.shape[
1] // 2 > (3 + 1) else dataset_train_y.shape[1]
range_labelset_size = list(range(3, end_range))
#starting_range = dataset_train_y.shape[1]//range_labelset_size[0]
range_model_count = list(
range(2 * dataset_train_y.shape[1],
2 * dataset_train_y.shape[1] + 1)
) #[x*2 for x in range((starting_range), (starting_range+1))]#[x*2 for x in range(dataset_train_y.shape[1]//6, dataset_train_y.shape[1]//2)]
print(dataset_train_y.shape[1])
print(range_labelset_size)
print(range_model_count)
parameters = [
{
'base_classifier': [GaussianNB()],
'labelset_size': range_labelset_size,
'model_count': range_model_count,
},
{
'base_classifier': [MultinomialNB()],
'base_classifier__alpha':
rangefloat, #for smoothing {Additive smoothing parameter NB}
'labelset_size': range_labelset_size,
'model_count': range_model_count,
},
{
'base_classifier': [SVC()],
'base_classifier__kernel': ['rbf', 'linear', 'sigmoid'],
'labelset_size': range_labelset_size,
'model_count': range_model_count,
},
]
classifier = GridSearchCV(classif,
parameters,
scoring=make_scorer(metrics.hamming_loss,
greater_is_better=False),
n_jobs=3)
classifier.fit(dataset_train_x, dataset_train_y)
return classifier.best_params_
def build_clf(X, y, top_estimator, base_estimator):
"""
Parameters
----------
X : NumPy Array
The array is generated by concatenating corpus and keywords vectors.
y : NumPy Array
y.shape[0] should be equal to X.shape[0].
top_estimator : str
One of ['ClassifierChain', 'LabelPowerset', 'Rake10'].
base_estimator : str
One of ['ClassifierChain', 'LabelPowerset', 'Rake10'].
In Scikit-learn, only the following support multilabel classification:
sklearn.tree.DecisionTreeClassifier
sklearn.tree.ExtraTreeClassifier
sklearn.ensemble.ExtraTreesClassifier
sklearn.neighbors.KNeighborsClassifier
sklearn.neural_network.MLPClassifier
sklearn.neighbors.RadiusNeighborsClassifier
sklearn.ensemble.RandomForestClassifier
sklearn.linear_model.RidgeClassifierCV
We have to use the `multiclass` or `multioutput` modules
if the base classifier is not among the above listed.
# In the multilabel learning literature,
# OvR is also known as the binary relevance method.
# An indicator matrix - a matrix of shape (n_samples, n_classes)
# turns on multilabel classification.
Returns
-------
A fitted classifier object.
"""
if base_estimator in ['KNeighbors', 'MLP', 'RandomForest']:
if base_estimator == 'KNeighbors':
"""
https://scikit-learn.org/stable/modules/generated/sklearn.neighbors.KNeighborsClassifier.html
"""
from sklearn.neighbors import KNeighborsClassifier
n_neighbors = int(input('Enter the number of neighbors\n'))
weights = input('Enter the weight function used in prediction\n')
leaf_size = int(input('Enter the leaf size\n'))
classifier = KNeighborsClassifier(n_neighbors=n_neighbors,
weights=weights,
leaf_size=leaf_size)
classifier.fit(X, y)
print('Training...\n')
return classifier
elif base_estimator == 'MLP':
"""
https://scikit-learn.org/stable/modules/generated/sklearn.neural_network.MLPClassifier.html
"""
from sklearn.neural_network import MLPClassifier
hidden_layer_sizes = tuple(eval(input('Enter the size of the hidden layers.\n e.g: (64,64,64)'\
'indicates a simple three-layer feedforward neural network with 64 neurons in each layer\n')))
classifier = MLPClassifier(
hidden_layer_sizes=hidden_layer_sizes,
solver='lbfgs') # use lbgfs because it's a small dataset
classifier.fit(X, y)
print('Training...\n')
return classifier
elif base_estimator == 'RandomForest':
"""
https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestClassifier.html
"""
from sklearn.ensemble import RandomForestClassifier
n_estimators = int(
input('Enter the number of trees in the forest.\n'))
max_depth = eval(
input('Enter the maximum depth of the tree (int or None)\n'))
min_samples_split = eval(
input(
'The minimum number of samples required to split an internal node\n'
))
min_samples_leaf = eval(
input(
'The minimum number of samples required to be at a leaf node\n'
))
classifier = RandomForestClassifier(
n_estimators=n_estimators,
max_depth=max_depth,
min_samples_split=min_samples_split,
min_samples_leaf=min_samples_leaf)
classifier.fit(X, y)
print('Training...\n')
return classifier
else:
if base_estimator == 'LogisticRegression':
"""
http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html
"""
from sklearn.linear_model import LogisticRegression
penalty = input(
'Enter the norm used in the penalization. One of {l1, l2, elasticnet, none}\n'
)
tol = float(
input(
'Enter the tolerance for stopping criteria; default=1e-4. Bigger number means faster convergence\n'
))
C = float(input('Enter the regularization strength. Smaller values specify stronger regularization.'\
'Must be a positive number.\n'))
solver = input(
'Enter the optimizer. One of {‘newton-cg’, ‘lbfgs’, ‘liblinear’, ‘sag’, ‘saga’}\n'
)
max_iter = int(input('Enter the maximum number of iterations.\n'))
base_classifier = LogisticRegression(penalty=penalty,
tol=tol,
C=C,
solver=solver,
max_iter=max_iter)
elif base_estimator == 'SGD':
"""
https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.SGDClassifier.html
"""
from sklearn.linear_model import SGDClassifier
loss = input(
'Enter the loss function. One of {‘hinge’, ‘log’, ‘modified_huber’, ‘squared_hinge’}\n'
)
penalty = input(
'Enter the penalty. One of {‘l2’, ‘l1’, ‘elasticnet’}\n')
tol = float(
input(
'Enter the tolerance for stopping criteria; default=1e-4. Bigger number means faster convergence\n'
))
learning_rate = input(
'Enter the learning rate: Either ‘optimal’ or ‘adaptive’\n')
eta0 = eval(input('Enter the initial learning rate\n'))
base_classifier = SGDClassifier(loss=loss,
penalty=penalty,
tol=tol,
learning_rate=learning_rate,
eta0=eta0,
alpha=0.0008)
if top_estimator == 'Rakelo':
classifier = RakelO(base_classifier=base_classifier,
base_classifier_require_dense=[True, True],
labelset_size=y.shape[1] // 4,
model_count=6)
classifier.fit(X, y)
print('Training...\n')
return classifier
else:
classifier = globals()[top_estimator](base_classifier)
classifier.fit(X, y)
print('Training...\n')
return classifier
def test_if_works_with_cross_validation(self):
classifier = RakelO(classifier=self.get_labelpowerset_with_nb(),
model_count=20,
labelset_size=5)
self.assertClassifierWorksWithCV(classifier)
def test_if_dense_classification_works_on_dense_base_classifier(self):
classifier = RakelO(classifier=self.get_labelpowerset_with_nb(),
model_count=20,
labelset_size=5)
self.assertClassifierWorksWithSparsity(classifier, 'dense')