def __init__(self,
smooth_iteration=25,
training_iteration=50,
spreading_factor=0.83,
FD=0.1,
learning_rate=0.3,
smooth_learning_factor=0.8):
"""
Called when initializing the classifier
"""
self.smooth_iteration = smooth_iteration
self.spreading_factor = spreading_factor
self.training_iteration = training_iteration
self.FD = FD
self.learning_rate = learning_rate
self.smooth_learning_factor = smooth_learning_factor
self.gsom = GSOM(self.spreading_factor,
55,
max_radius=4,
FD=self.FD,
learning_rate=self.learning_rate,
smooth_learning_factor=self.smooth_learning_factor)
self.gsmote = GeometricSMOTE(random_state=1,
truncation_factor=1.0,
deformation_factor=0,
k_neighbors=5,
sampling_rate=0.3)
def OverSample(X, y):
print('Original dataset shape %s' % Counter(y))
gsmote = GeometricSMOTE(random_state=1)
X_res, y_res = gsmote.fit_resample(X, y)
print('Resampled dataset shape %s' % Counter(y_res))
return X_res, y_res
def __init__(self,
truncation_factor=1.0,
deformation_factor=0.0,
k_neighbors=1,
sampling_rate=0.3,
n_estimators=100,
learning_rate=0.01,
max_depth=3):
"""
Called when initializing the classifier
"""
self.truncation_factor = truncation_factor
self.deformation_factor = deformation_factor
self.k_neighbors = k_neighbors
self.sampling_rate = sampling_rate
self.n_estimators = n_estimators
self.learning_rate = learning_rate
self.max_depth = max_depth
self.regressor = GradientBoostingClassifier(
n_estimators=self.n_estimators,
learning_rate=self.learning_rate,
max_depth=self.max_depth)
self.gsmote = GeometricSMOTE(
random_state=1,
truncation_factor=self.truncation_factor,
deformation_factor=self.deformation_factor,
k_neighbors=self.k_neighbors,
sampling_rate=self.sampling_rate)
def generate_oversamplers(factor):
"""Generate a list of oversamplers that pre-apply undersampling."""
if factor is None:
return [('BENCHMARK METHOD', None, {})]
return [('NO OVERSAMPLING',
UnderOverSampler(oversampler=None, factor=factor), {}),
('RANDOM OVERSAMPLING',
UnderOverSampler(oversampler=RandomOverSampler(),
factor=factor), {}),
('SMOTE', UnderOverSampler(oversampler=SMOTE(), factor=factor), {
'oversampler__k_neighbors': [3, 5]
}),
('BORDERLINE SMOTE',
UnderOverSampler(oversampler=BorderlineSMOTE(), factor=factor), {
'oversampler__k_neighbors': [3, 5]
}),
('G-SMOTE',
UnderOverSampler(oversampler=GeometricSMOTE(), factor=factor), {
'oversampler__k_neighbors': [3, 5],
'oversampler__selection_strategy':
['combined', 'minority', 'majority'],
'oversampler__truncation_factor':
[-1.0, -0.5, .0, 0.25, 0.5, 0.75, 1.0],
'oversampler__deformation_factor':
[.0, 0.2, 0.4, 0.5, 0.6, 0.8, 1.0]
})]
class MeanClassifier(BaseEstimator, ClassifierMixin):
"""An example of classifier"""
def __init__(self,
smooth_iteration=25,
training_iteration=50,
spreading_factor=0.83,
FD=0.1,
learning_rate=0.3,
smooth_learning_factor=0.8):
"""
Called when initializing the classifier
"""
self.smooth_iteration = smooth_iteration
self.spreading_factor = spreading_factor
self.training_iteration = training_iteration
self.FD = FD
self.learning_rate = learning_rate
self.smooth_learning_factor = smooth_learning_factor
self.gsom = GSOM(self.spreading_factor,
55,
max_radius=4,
FD=self.FD,
learning_rate=self.learning_rate,
smooth_learning_factor=self.smooth_learning_factor)
self.gsmote = GeometricSMOTE(random_state=1,
truncation_factor=1.0,
deformation_factor=0,
k_neighbors=5,
sampling_rate=0.3)
def fit(self, X, y):
X_train, y_train = self.gsmote.fit_resample(X, y)
y1 = np.copy(y_train)
y = np.column_stack([y1, y_train])
labels = ["Name", "label"]
y = np.vstack((labels, y))
frame = pd.DataFrame(y[1:, :], columns=y[0, :])
self.gsom.fit(X_train, self.training_iteration, self.smooth_iteration)
self.gsom.labelling_gsom(X_train, frame, "Name", "label")
self.gsom.finalize_gsom_label()
return self
# def _meaning(self, x):
# return True
def predict(self, X):
return self.gsom.predict_values(X)
class MeanClassifier(BaseEstimator, ClassifierMixin):
"""An example of classifier"""
def __init__(self,
truncation_factor=1.0,
deformation_factor=0.0,
k_neighbors=1,
sampling_rate=0.3,
n_estimators=100,
learning_rate=0.01,
max_depth=3):
"""
Called when initializing the classifier
"""
self.truncation_factor = truncation_factor
self.deformation_factor = deformation_factor
self.k_neighbors = k_neighbors
self.sampling_rate = sampling_rate
self.n_estimators = n_estimators
self.learning_rate = learning_rate
self.max_depth = max_depth
self.regressor = GradientBoostingClassifier(
n_estimators=self.n_estimators,
learning_rate=self.learning_rate,
max_depth=self.max_depth)
self.gsmote = GeometricSMOTE(
random_state=1,
truncation_factor=self.truncation_factor,
deformation_factor=self.deformation_factor,
k_neighbors=self.k_neighbors,
sampling_rate=self.sampling_rate)
def fit(self, X, y):
print(self.max_depth, self.learning_rate, self.n_estimators,
self.sampling_rate, self.k_neighbors, self.deformation_factor,
self.truncation_factor)
X_train, y_train = self.gsmote.fit_resample(X, y)
self.regressor.fit(X_train, y_train)
return self
# def _meaning(self, x):
# return True
def predict(self, y):
return self.regressor.predict(y)
def model_fit(X_train, y_train, X_test, y_test):
classifier_dict = {
"no_oversampling":
Pipeline([("none", None),
("lr", LogisticRegression(solver="liblinear"))]),
"smote":
Pipeline([
("smote", SMOTE(k_neighbors=1)),
("lr", LogisticRegression(solver="liblinear")),
]),
"gsmote":
Pipeline([
("gsmote", GeometricSMOTE(k_neighbors=1)),
("lr", LogisticRegression(solver="liblinear")),
]),
}
results = {}
for name, estimator in classifier_dict.items():
estimator.fit(X_train, y_train)
results[name] = estimator.score(X_test, y_test)
return pd.DataFrame(data=results.values(),
index=results.keys(),
columns=["score"])
plt.ylim([0.9, 1.0])
###############################################################################
# Low Imbalance Ratio or high Samples to Features Ratio
###############################################################################
###############################################################################
# When :math:`\text{IR} = \frac{\text{\# majority samples}}{\text{\# minority
# samples}}` (Imbalance Ratio) is low or :math:`\text{SFR} = \frac{\text{\#
# samples}}{\text{\# features}}` (Samples to Features Ratio) is high then the
# minority selection strategy and higher absolute values of the truncation and
# deformation factors dominate as optimal hyperparameters.
X, y = generate_imbalanced_data([0.3, 0.7], 2000, 6, 4)
gsmote_gbc = make_pipeline(GeometricSMOTE(random_state=RANDOM_STATE),
DecisionTreeClassifier(random_state=RANDOM_STATE))
scoring_name = 'Geometric Mean Score'
validation_curve_info = generate_validation_curve_info(gsmote_gbc, X, y, range(1, 8), "geometricsmote__k_neighbors", SCORER)
plot_validation_curve(validation_curve_info, scoring_name, 'K Neighbors')
validation_curve_info = generate_validation_curve_info(gsmote_gbc, X, y, np.linspace(-1.0, 1.0, 9), "geometricsmote__truncation_factor", SCORER)
plot_validation_curve(validation_curve_info, scoring_name, 'Truncation Factor')
validation_curve_info = generate_validation_curve_info(gsmote_gbc, X, y, np.linspace(0.0, 1.0, 5), "geometricsmote__deformation_factor", SCORER)
plot_validation_curve(validation_curve_info, scoring_name, 'Deformation Factor')
validation_curve_info = generate_validation_curve_info(gsmote_gbc, X, y, ['minority', 'majority', 'combined'], "geometricsmote__selection_strategy", SCORER)
plot_validation_curve(validation_curve_info, scoring_name, 'Selection Strategy')
def runSMOTEvariationsGen(self, folder):
"""
Create files with SMOTE preprocessing and without preprocessing.
:param datasets: datasets.
:param folder: cross-validation folders.
:return:
"""
smote = SMOTE()
borderline1 = BorderlineSMOTE(kind='borderline-1')
borderline2 = BorderlineSMOTE(kind='borderline-2')
smoteSVM = SVMSMOTE()
geometric_smote = GeometricSMOTE(n_jobs=-1)
for dataset in datasets: # biclass e multiclass
for fold in range(5):
path = os.path.join(folder, dataset, str(fold),
''.join([dataset, "_train.csv"]))
train = np.genfromtxt(path, delimiter=',')
X = train[:, 0:train.shape[1] - 1]
Y = train[:, train.shape[1] - 1]
# SMOTE
print("SMOTE..." + dataset)
X_res, y_res = smote.fit_sample(X, Y)
y_res = y_res.reshape(len(y_res), 1)
newdata = np.hstack([X_res, y_res])
newtrain = pd.DataFrame(newdata)
newtrain.to_csv(os.path.join(folder, dataset, str(fold),
''.join([dataset, "_SMOTE.csv"])),
header=False,
index=False)
# SMOTE BORDERLINE1
print("Borderline1..." + dataset)
X_res, y_res = borderline1.fit_sample(X, Y)
y_res = y_res.reshape(len(y_res), 1)
newdata = np.hstack([X_res, y_res])
newtrain = pd.DataFrame(newdata)
newtrain.to_csv(os.path.join(
folder, dataset, str(fold),
''.join([dataset, "_Borderline1.csv"])),
header=False,
index=False)
# SMOTE BORDERLINE2
print("Borderline2..." + dataset)
X_res, y_res = borderline2.fit_sample(X, Y)
y_res = y_res.reshape(len(y_res), 1)
newdata = np.hstack([X_res, y_res])
newtrain = pd.DataFrame(newdata)
newtrain.to_csv(os.path.join(
folder, dataset, str(fold),
''.join([dataset, "_Borderline2.csv"])),
header=False,
index=False)
# SMOTE SVM
print("SMOTE SVM..." + dataset)
X_res, y_res = smoteSVM.fit_sample(X, Y)
y_res = y_res.reshape(len(y_res), 1)
newdata = np.hstack([X_res, y_res])
newtrain = pd.DataFrame(newdata)
newtrain.to_csv(os.path.join(
folder, dataset, str(fold),
''.join([dataset, "_smoteSVM.csv"])),
header=False,
index=False)
# GEOMETRIC SMOTE
print("GEOMETRIC SMOTE..." + dataset)
X_res, y_res = geometric_smote.fit_resample(X, Y)
y_res = y_res.reshape(len(y_res), 1)
newdata = np.hstack([X_res, y_res])
newtrain = pd.DataFrame(newdata)
newtrain.to_csv(os.path.join(
folder, dataset, str(fold),
''.join([dataset, "_Geometric_SMOTE.csv"])),
header=False,
index=False)
def parse_input_zoo_data(filename, header='infer'):
gsmote = GeometricSMOTE(random_state=1)
#
# (X_train, y_train), (X_test, y_test) = mnist.load_data()
#
# d1, d2, d3 = X_train.shape
# X_train_reshaped = X_train.reshape(d1, d2 * d3)
# print(X_train_reshaped[:2000, :].shape)
# y_train_half = y_train[:2000]
# classes = y_train_half.tolist()
# labels = y_train_half.tolist()
# # print(labels)
#
# input_database = {
# 0: X_train_reshaped[:2000, :]
# }
#GSMOTE
# X_f,y_f = GSMOTE.OverSample()
#
#
# X_t, X_test, y_t, y_test = train_test_split(X_f, y_f, test_size=0.2, random_state=0)
#
#
# classes = y_t.tolist()
# labels = y_t.tolist()
# input_database = {
# 0: X_t
# }
X, y = pp.preProcess(filename)
X_t, X_test, y_t, y_test = train_test_split(X,
y,
test_size=0.2,
random_state=0)
X_train, y_train = gsmote.fit_resample(X_t, y_t)
classes = y_train.tolist()
labels = y_train.tolist()
input_database = {0: X_train}
# (X_train, y_train), (X_test, y_test) = mnist.load_data()
#
# d1, d2, d3 = X_train.shape
# X_train_reshaped = X_train.reshape(d1, d2 * d3)
# print(X_train_reshaped[:2000, :].shape)
# y_train_half = y_train[:2000]
# classes = y_train_half.tolist()
# labels = y_train_half.tolist()
# # print(labels)
#
# input_database = {
# 0: X_train_reshaped[:2000, :]
# }
#Smote
# X_f,y_f = smote.Data_Extract(filename)
# classes = y_f.tolist()
# labels = y_f.tolist()
# input_database = {
# 0: X_f[:,:]
# }
# input_data = pd.read_csv(filename, header=header)
#
# input_database = {
# 0: input_data.as_matrix([0,1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16,17,18,19,20,21,22,23,24,25,26,27,28,29])
# }
#
# (X_train, y_train), (X_test, y_test) = mnist.load_data()
#
# d1, d2, d3 = X_train.shape
# X_train_reshaped = X_train.reshape(d1, d2 * d3)
# print(X_train_reshaped[:2000, :].shape)
# y_train_half = y_train[:2000]
# classes = y_train_half.tolist()
# labels = y_train_half.tolist()
# # print(labels)
#
# input_database = {
# 0: X_train_reshaped[:2000, :]
# }
# input_data = pd.read_csv(filename, header=header)
#
# classes = input_data[17].tolist()
# labels = input_data[0].tolist()
# input_database = {
# 0: input_data.as_matrix([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16])
# }
return input_database, labels, classes, X_test, y_test
###############################################################################
# Truncation factor
#..............................................................................
#
# The hyperparameter ``truncation_factor`` determines the degree of truncation
# that is applied on the initial geometric area. Selecting the values of
# geometric hyperparameters as `truncation_factor=0.0`,
# ``deformation_factor=0.0`` and ``selection_strategy='minority'``, the data
# generation area in 2D corresponds to a circle with center as one of the two
# minority class samples and radius equal to the distance between them. In the
# multi-dimensional case the corresponding area is a hypersphere. When
# truncation factor is increased, the hypersphere is truncated and for
# ``truncation_factor=1.0`` becomes a half-hypersphere. Negative values of
# ``truncation_factor`` have a similar effect but on the opposite direction.
gsmote = GeometricSMOTE(k_neighbors=1, deformation_factor=0.0,
selection_strategy='minority', random_state=RANDOM_STATE)
truncation_factors = np.array([0.0, 0.2, 0.4, 0.6, 0.8, 1.0])
n_subplots = [2, 3]
plot_hyperparameters(gsmote, X, y, 'truncation_factor', truncation_factors, n_subplots)
plot_hyperparameters(gsmote, X, y, 'truncation_factor', -truncation_factors, n_subplots)
###############################################################################
# Deformation factor
#..............................................................................
#
# When the ``deformation_factor`` is increased, the data generation area deforms
# to an ellipsis and for ``deformation_factor=1.0`` becomes a line segment.
gsmote = GeometricSMOTE(k_neighbors=1, truncation_factor=0.0,
selection_strategy='minority', random_state=RANDOM_STATE)
deformation_factors = np.array([0.0, 0.2, 0.4, 0.6, 0.8, 1.0])
#
# Below we use the Random Forest Classifier to predict the forest type of each
# patch of forest. Two experiments are ran: One using only the classifier and
# another that creates a pipeline of Geometric SMOTE and the classifier. A
# classification report is printed for both experiments.
splitted_data = train_test_split(X,
y,
test_size=0.95,
random_state=RANDOM_STATE,
shuffle=True)
clf = RandomForestClassifier(bootstrap=True,
n_estimators=10,
random_state=RANDOM_STATE)
ovs_clf = make_pipeline(GeometricSMOTE(random_state=RANDOM_STATE), clf)
print_classification_report(clf, *splitted_data)
print_classification_report(ovs_clf, *splitted_data)
###############################################################################
# Indian Pines
###############################################################################
###############################################################################
# This hyperspectral data set has 220 spectral bands and 20 m spatial resolution.
# The imagery was collected on 12 June 1992 and represents a 2.9 by 2.9 km area
# in Tippecanoe County, Indiana, USA. The area is agricultural and eight classes
# as land-use types are presented: alfalfa, corn, grass, hay, oats, soybeans,
# trees, and wheat. The Indian Pines data set has been used for testing and
# comparing algorithms. The number of samples varies greatly among the classes,
def _check_estimators(self, X, y):
"""Check various estimators."""
# Import SOM and GeometricSMOTE
try:
from somlearn import SOM
except ImportError:
raise ImportError(
'GeometricSOMO class requires the package `som-learn` to be installed.'
)
try:
from gsmote import GeometricSMOTE
except ImportError:
raise ImportError(
'GeometricSOMO class requires the package `geometric-smote` to '
'be installed.')
# Check oversampler
self.oversampler_ = GeometricSMOTE(
sampling_strategy=self.sampling_strategy,
k_neighbors=self.k_neighbors,
truncation_factor=self.truncation_factor,
deformation_factor=self.deformation_factor,
selection_strategy=self.selection_strategy,
random_state=self.random_state_,
n_jobs=self.n_jobs,
)
if self.som_estimator is None:
self.clusterer_ = SOM(random_state=self.random_state_)
elif isinstance(self.som_estimator, int):
check_scalar(self.som_estimator, 'som_estimator', int, min_val=1)
n = round(sqrt(self.som_estimator))
self.clusterer_ = SOM(n_columns=n,
n_rows=n,
random_state=self.random_state_)
elif isinstance(self.som_estimator, float):
check_scalar(self.som_estimator,
'som_estimator',
float,
min_val=0.0,
max_val=1.0)
n = round(sqrt((X.shape[0] - 1) * self.som_estimator + 1))
self.clusterer_ = SOM(n_columns=n,
n_rows=n,
random_state=self.random_state_)
elif isinstance(self.som_estimator, SOM):
self.clusterer_ = clone(self.som_estimator)
else:
raise TypeError('Parameter `som_estimator` should be '
'either `None` or the number of clusters '
'or a float in the [0.0, 1.0] range equal to'
' the number of clusters over the number of '
'samples or an instance of the `SOM` class.')
# Check distributor
self.distributor_ = DensityDistributor(
filtering_threshold=self.imbalance_ratio_threshold,
distances_exponent=self.distances_exponent,
distribution_ratio=self.distribution_ratio,
)
return self
def load_best_classifier(self, X, y):
scores = []
for _, classifier in self.classifiers_:
scores.append(self.evaluation_metric_(classifier, X, y))
self.classifier_ = self.classifiers_[np.argmax(scores)][-1]
return self
def predict(self, X):
return self.classifier_.predict(X)
CONFIG = {
'oversamplers': [
#('NONE', None, {}),
('G-SMOTE', ClusterOverSampler(GeometricSMOTE(), n_jobs=1), {})
# 'oversampler__k_neighbors': [3, 5],
# 'oversampler__selection_strategy': ['combined', 'minority', 'majority'],
# 'oversampler__truncation_factor': [-1.0, .0, 1.0],
# 'oversampler__deformation_factor': [.0, 0.5, 1.0]
# })
],
'classifiers': [
('LR',
LogisticRegression(multi_class='multinomial',
solver='sag',
penalty='none',
max_iter=1e4), {}),
('KNN', KNeighborsClassifier(), {
'n_neighbors': [3]
}), #, 5, 8]}),
###############################################################################
# Low Imbalance Ratio or high Samples to Features Ratio
###############################################################################
###############################################################################
# When :math:`\text{IR} = \frac{\text{\# majority samples}}{\text{\# minority
# samples}}` (Imbalance Ratio) is low or :math:`\text{SFR} = \frac{\text{\#
# samples}}{\text{\# features}}` (Samples to Features Ratio) is high then the
# minority selection strategy and higher absolute values of the truncation and
# deformation factors dominate as optimal hyperparameters.
X, y = generate_imbalanced_data([0.3, 0.7], 2000, 6, 4)
gsmote_gbc = make_pipeline(
GeometricSMOTE(random_state=RANDOM_STATE),
DecisionTreeClassifier(random_state=RANDOM_STATE),
)
scoring_name = 'Geometric Mean Score'
validation_curve_info = generate_validation_curve_info(
gsmote_gbc, X, y, range(1, 8), "geometricsmote__k_neighbors", SCORER)
plot_validation_curve(validation_curve_info, scoring_name, 'K Neighbors')
validation_curve_info = generate_validation_curve_info(
gsmote_gbc,
X,
y,
np.linspace(-1.0, 1.0, 9),
"geometricsmote__truncation_factor",
SCORER,
'learning_rate': [0.01],
'max_depth': [3]
}]
gs = GridSearchCV(MeanClassifier(), parameters)
gs.fit(X, y)
params = gs.best_params_
print(params)
#find performance
X_t, X_test, y_t, y_test = train_test_split(X,
y,
test_size=0.2,
random_state=0)
gsmote = GeometricSMOTE(random_state=1,
truncation_factor=params["truncation_factor"],
deformation_factor=params["deformation_factor"],
k_neighbors=params["k_neighbors"],
sampling_rate=params["sampling_rate"])
X_train, y_train = gsmote.fit_resample(X_t, y_t)
# Fitting Gradient boosting
gbc = GradientBoostingClassifier(n_estimators=params["n_estimators"],
learning_rate=params["learning_rate"],
max_depth=params["max_depth"])
gbc.fit(X_train, y_train)
# Predicting the Test set results
y_predict = gbc.predict(X_test)
y_pred = np.where(y_predict.astype(int) > 0.5, 1, 0)
evaluate("Gradient Boosting", y_test, y_pred)
X_largest, y_largest, X_smallest, y_smallest = X2, y2, X1, y1
intersecting_vals = np.in1d(X_largest, X_smallest).reshape(X_largest.shape)
disjoin_indexes = np.where(~np.all(intersecting_vals, axis=1))[0]
return X_largest.iloc[disjoin_indexes], y_largest.iloc[disjoin_indexes]
for strategy in ["combined", "majority", "minority"]:
X_gsmote_final = np.empty(shape=(0, X_train.shape[-1]))
y_gsmote_final = np.empty(shape=(0))
for d in [0, 0.5, 1]:
for t in [-1, 0, 1]:
gsmote_sampling = GeometricSMOTE(
k_neighbors=1,
deformation_factor=d,
truncation_factor=t,
n_jobs=-1,
selection_strategy=strategy,
).fit_resample(X_train, y_train)
X_gsmote, _ = get_disjoin(X_train, y_train, gsmote_sampling[0],
gsmote_sampling[1])
X_gsmote_final = np.append(X_gsmote_final, X_gsmote, axis=0)
y_gsmote_final = np.append(y_gsmote_final,
np.array([f"t={t}, d={d}"] *
X_gsmote.shape[0]),
axis=0)
plot_mnist_samples(
pd.DataFrame(X_gsmote_final),
pd.Series(y_gsmote_final),
f"Generated Using G-SMOTE: {strategy}",
)
"""Geometric mean score with macro average."""
return geometric_mean_score(y_true, y_pred, average='macro')
SCORERS['geometric_mean_score_macro'] = make_scorer(geometric_mean_score_macro)
CONFIG = {
'oversamplers':
[('NONE', None, {}), ('ROS', RandomOverSampler(), {}),
('SMOTE', SMOTE(), {
'k_neighbors': [3, 5]
}), ('B-SMOTE', BorderlineSMOTE(), {
'k_neighbors': [3, 5]
}), ('ADASYN', ADASYN(), {
'n_neighbors': [2, 3]
}),
('G-SMOTE', GeometricSMOTE(), {
'k_neighbors': [3, 5],
'selection_strategy': ['combined', 'minority', 'majority'],
'truncation_factor': [-1.0, -0.5, .0, 0.25, 0.5, 0.75, 1.0],
'deformation_factor': [.0, 0.2, 0.4, 0.5, 0.6, 0.8, 1.0]
})],
'classifiers': [
('CONSTANT CLASSIFIER', DummyClassifier(strategy='constant',
constant=0), {}),
('LR', LogisticRegression(solver='liblinear', multi_class='auto'), {}),
('KNN', KNeighborsClassifier(), {
'n_neighbors': [3, 5]
}), ('DT', DecisionTreeClassifier(), {
'max_depth': [3, 6]
}),
('GBC', GradientBoostingClassifier(), {
def runSMOTEvariationsGen(self, folder):
"""
Create files with SMOTE preprocessing and without preprocessing.
:param datasets: datasets.
:param folder: cross-validation folders.
:return:
"""
smote = SMOTE()
borderline1 = BorderlineSMOTE(kind='borderline-1')
borderline2 = BorderlineSMOTE(kind='borderline-2')
smoteSVM = SVMSMOTE()
geometric_smote = GeometricSMOTE(n_jobs=-1)
for dataset in datasets:
for fold in range(5):
path = os.path.join(folder, dataset, str(fold), ''.join([dataset, "_train.csv"]))
train = np.genfromtxt(path, delimiter=',')
X = train[:, 0:train.shape[1] - 1]
Y = train[:, train.shape[1] - 1]
Y = Y.reshape(len(Y), 1)
# SMOTE
print("SMOTE..." + dataset)
data_r = np.hstack([X, Y])
data_r = pd.DataFrame(data_r)
data_r.columns = data_r.columns.astype(str)
colunas = list(data_r.columns)
y_name = colunas[-1]
dtoregression = dtosmoter(
data=data_r,
y=y_name,
oversampler=smote
)
dtoregression.to_csv(os.path.join(folder, dataset, str(fold), ''.join([dataset, "_SMOTE.csv"])),
header=False, index=False)
# SMOTE BORDERLINE1
print("Borderline1..." + dataset)
data_r = np.hstack([X, Y])
data_r = pd.DataFrame(data_r)
data_r.columns = data_r.columns.astype(str)
colunas = list(data_r.columns)
y_name = colunas[-1]
dtoregression = dtosmoter(
data=data_r,
y=y_name,
oversampler=borderline1
)
dtoregression.to_csv(os.path.join(folder, dataset, str(fold), ''.join([dataset, "_Borderline1.csv"])),
header=False, index=False)
# SMOTE BORDERLINE2
print("Borderline2..." + dataset)
data_r = np.hstack([X, Y])
data_r = pd.DataFrame(data_r)
data_r.columns = data_r.columns.astype(str)
colunas = list(data_r.columns)
y_name = colunas[-1]
dtoregression = dtosmoter(
data=data_r,
y=y_name,
oversampler=borderline2
)
dtoregression.to_csv(os.path.join(folder, dataset, str(fold), ''.join([dataset, "_Borderline2.csv"])),
header=False, index=False)
# SMOTE SVM
print("SMOTE SVM..." + dataset)
data_r = np.hstack([X, Y])
data_r = pd.DataFrame(data_r)
data_r.columns = data_r.columns.astype(str)
colunas = list(data_r.columns)
y_name = colunas[-1]
dtoregression = dtosmoter(
data=data_r,
y=y_name,
oversampler=smoteSVM
)
dtoregression.to_csv(os.path.join(folder, dataset, str(fold), ''.join([dataset, "_smoteSVM.csv"])),
header=False, index=False)
# GEOMETRIC SMOTE
print("GEOMETRIC SMOTE..." + dataset)
data_r = np.hstack([X, Y])
data_r = pd.DataFrame(data_r)
data_r.columns = data_r.columns.astype(str)
colunas = list(data_r.columns)
y_name = colunas[-1]
dtoregression = dtosmoter(
data=data_r,
y=y_name,
oversampler=geometric_smote
)
dtoregression.to_csv(
os.path.join(folder, dataset, str(fold), ''.join([dataset, "_Geometric_SMOTE.csv"])),
header=False, index=False)
