def _oversample(X, y, method='SMOTE', strat='not majority'):
# compute minimum number of samples per class
min_samples = len(y)
for l in set(y):
if y.tolist().count(l) < min_samples:
min_samples = y.tolist().count(l)
if min_samples <= 5:
method = 'RNDM'
if method == 'ADASYN':
ios = imbover.ADASYN(sampling_strategy=strat, random_state=42)
elif method == 'SMOTE':
ios = imbover.SMOTE(sampling_strategy=strat, random_state=42)
elif method == 'SMOTENC':
ios = imbover.SMOTENC(sampling_strategy=strat, random_state=42)
elif method == 'BORDERSMOTE':
ios = imbover.BorderlineSMOTE(sampling_strategy=strat, random_state=42)
elif method == 'SVMSMOTE':
ios = imbover.SVMSMOTE(sampling_strategy=strat, random_state=42)
elif method == 'KMEANSSMOTE':
ios = imbover.KMeansSMOTE(sampling_strategy=strat, random_state=42)
elif method == 'RNDM':
ios = imbover.RandomOverSampler(sampling_strategy=strat,
random_state=42)
X_resampled, y_resampled = ios.fit_resample(X, y)
return X_resampled, y_resampled
def __init__(self,
inputs,
targets,
batch_size=100,
max_num_batches=-1,
shuffle_order=True,
rng=None,
oversample=None):
"""Create a new recognition data provider object.
Args:
inputs (ndarray): Array of data input features of shape
(num_data, input_dim).
targets (ndarray): Array of data output targets of shape
(num_data, output_dim) or (num_data,) if output_dim == 1.
batch_size (int): Number of data points to include in each batch.
max_num_batches (int): Maximum number of batches to iterate over
in an epoch. If `max_num_batches * batch_size > num_data` then
only as many batches as the data can be split into will be
used. If set to -1 all of the data will be used.
shuffle_order (bool): Whether to randomly permute the order of
the data before each epoch.
rng (RandomState): A seeded random number generator.
"""
if not oversample is None:
oversample = oversample.lower()
self.initialize_seed(rng)
if oversample == "smote":
oversampler = imbl.SMOTE(random_state=self.rng)
elif oversample == "smote-cat":
# Need method for specifying categorical attributes, e.g., imbl.SMOTENC(random_state=self.rng, categorical_features=range(4200, 4348))
raise (NotImplementedError)
elif oversample == "smote-svm":
oversampler = imbl.SVMSMOTE(random_state=self.rng)
elif oversample == "smote-borderline-1":
oversampler = imbl.BorderlineSMOTE(random_state=self.rng,
kind='borderline-1')
elif oversample == "smote-borderline-2":
oversampler = imbl.BorderlineSMOTE(random_state=self.rng,
kind='borderline-2')
elif oversample == "adasyn":
oversampler = imbl.ADASYN(random_state=self.rng)
else:
raise (Exception(
"Unrecognized oversampling method: {0}".format(oversample))
)
inputs, targets = oversampler.fit_resample(inputs, targets)
self.num_classes = 3
inputs = inputs.astype(np.float32)
# pass the loaded data to the parent class __init__
super(RecognitionDataProvider,
self).__init__(inputs, targets, batch_size, max_num_batches,
shuffle_order, rng)
feature_src[where_are_inf] = 0
feature_src = mami.fit_transform(feature_src)
where_are_nan = np.isnan(feature_tar)
where_are_inf = np.isinf(feature_tar)
feature_tar[where_are_nan] = 0
feature_tar[where_are_inf] = 0
feature_tar = mami.fit_transform(feature_tar)
train_label = train_label.values
src_label = train_label[:, 0]
test_label = test_label.values
tar_label = test_label[:, 0]
print('Before resampled dataset shape %s' % Counter(src_label))
sm = over_sampling.SVMSMOTE(sampling_strategy={1: 860, 0: 600})
# feature_src, src_label = sm.fit_resample(feature_src, src_label)
# print('Resampled dataset shape %s' % Counter(src_label))
from sklearn.linear_model import Lasso, LassoCV
""" Lasso 400 features 0.000005 """
lasso = Lasso(alpha=0.0005)
lasso.fit(feature_src, src_label)
coef = lasso.coef_
coef.astype('float64')
pos = np.where(abs(coef) > 0.000001)
pos = np.array(pos)
pos = pos[0, :]
print(len(pos))
def svmsmote(features, labels):
smote = over_sampling.SVMSMOTE(random_state=0)
return smote.fit_resample(X=features, y=labels)
y_train, y_test = y_data.iloc[train_index], y_data.iloc[test_index]
features = X_train.columns.to_list()
# imputation
imp = SimpleImputer(missing_values=np.nan, strategy='median')
imp.fit(X_train)
X_train = imp.transform(X_train)
X_test = imp.transform(X_test)
# scaling
scaler = preprocessing.MinMaxScaler()
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)
# over-sampling
# print('before', y_train.groupby(['Label']).size())
sm = over_sampling.SVMSMOTE(random_state=42)
X_train, y_train = sm.fit_resample(X_train, y_train)
# feature selection
model = ExtraTreesClassifier(n_estimators=250, random_state=42)
model.fit(X_train, y_train.values.ravel())
importances = model.feature_importances_
no_zero_importance = importances[np.where(importances > 0)]
cutoff = np.std(no_zero_importance) + np.min(no_zero_importance)
indices = np.where(importances > cutoff)[0]
fs_elements = np.concatenate((fs_elements, np.array(features)[indices]),
axis=0)
# for i in indices:
# print(features[i], importances[i])
X_train_fs = X_train[:, indices]
X_test_fs = X_test[:, indices]
def main(X=None, y=None): # ndarray
fig, axes = plt.subplots(4, 2, figsize=(10, 10))
X = pd.DataFrame(X)
print('\nOriginal dataset shape %s' % Counter(y), '\n')
font2 = {
'family': 'Times New Roman',
'size': 20,
}
'''weight-smote'''
sm = all_smote_v2.SMOTE(random_state=42)
X_res, y_res = sm.fit_resample(X, y)
print('weight smote %s' % Counter(y_res), '\n\n\n')
y_res = pd.DataFrame(y_res)
X_new = X_res.iloc[len(X):, :]
y_new = y_res.iloc[len(y):, :]
for i in range(len(X)):
if y[i] == 1:
# print(X.iloc[i][0])
axes[0][0].scatter(X.iloc[i][0], X.iloc[i][1], c='tan', s=25)
for i in range(len(X)):
if y[i] == 0:
axes[0][0].scatter(X.iloc[i][0], X.iloc[i][1], c='darkcyan', s=25)
axes[0][0].scatter(X_new[0], X_new[1], c='red', marker='+', s=50)
axes[0][0].set_title('(a) weight-SMOTE ', font2)
'''smote'''
sm = over_sampling.SMOTE(random_state=42)
X_res, y_res = sm.fit_resample(X, y)
print('smote %s' % Counter(y_res), '\n\n\n')
y_res = pd.DataFrame(y_res)
X_new = X_res.iloc[len(X):, :]
y_new = y_res.iloc[len(y):, :]
for i in range(len(X)):
if y[i] == 1:
# print(X.iloc[i][0])
axes[0][1].scatter(X.iloc[i][0], X.iloc[i][1], c='tan', s=25)
for i in range(len(X)):
if y[i] == 0:
axes[0][1].scatter(X.iloc[i][0], X.iloc[i][1], c='darkcyan', s=25)
axes[0][1].scatter(X_new[0], X_new[1], c='red', marker='+', s=50)
axes[0][1].set_title('(b) SMOTE ', font2)
'''boderline_1'''
sm_1 = over_sampling.BorderlineSMOTE(random_state=42, kind="borderline-1")
X_res, y_res = sm_1.fit_resample(X, y)
print('borderline-1 shape %s' % Counter(y_res), '\n\n\n\n')
y_res = pd.DataFrame(y_res)
X_new = X_res.iloc[len(X):, :]
y_new = y_res.iloc[len(y):, :]
for i in range(len(X)):
if y[i] == 1:
# print(X.iloc[i][0])
axes[1][1].scatter(X.iloc[i][0], X.iloc[i][1], c='tan', s=25)
for i in range(len(X)):
if y[i] == 0:
axes[1][1].scatter(X.iloc[i][0], X.iloc[i][1], c='darkcyan', s=25)
# axes[0][0].scatter(X.iloc[i:,0:], X[i:,1:], c='cyan',)
axes[1][1].scatter(X_new[0], X_new[1], c='red', marker='+', s=50)
axes[1][1].set_title('(d) boderline-SMOTE1', font2)
'''weight-boderline'''
sm_zhou = all_smote_v2.BorderlineSMOTE(
random_state=42,
kind="weight-borderline-smote",
)
X_res, y_res = sm_zhou.fit_resample(X, y)
print('weight-borderline shape %s' % Counter(y_res), '\n\n\n')
y_res = pd.DataFrame(y_res)
X_new = X_res.iloc[len(X):, :]
y_new = y_res.iloc[len(y):, :]
for i in range(len(X)):
if y[i] == 1:
# print(X.iloc[i][0])
axes[1][0].scatter(X.iloc[i][0], X.iloc[i][1], c='tan', s=25)
for i in range(len(X)):
if y[i] == 0:
axes[1][0].scatter(X.iloc[i][0], X.iloc[i][1], c='darkcyan', s=25)
# axes[0][0].scatter(X.iloc[i:,0:], X[i:,1:], c='cyan',)
axes[1][0].scatter(X_new[0], X_new[1], c='red', marker='+', s=50)
axes[1][0].set_title('(c) weight-boderline', font2)
'''weight-kmeans-smote'''
sm_3 = all_smote_v2.KMeansSMOTE(random_state=42, kind='kmeans-borderline')
X_res, y_res = sm_3.fit_resample(X, y)
print('weight-kmeans-smote:\t', Counter(y_res), '\n\n\n\n')
y_res = pd.DataFrame(y_res)
X_new = X_res.iloc[len(X):, :]
y_new = y_res.iloc[len(y):, :]
axes[2][0].scatter(X[0], X[1], c=y, alpha=0.5)
axes[2][0].scatter(X_new[0], X_new[1], c='red', alpha=0.2)
axes[2][0].set_title('weight_kmeans_smote')
'''kmeans-smote'''
sm_4 = over_sampling.KMeansSMOTE(random_state=42, )
X_res, y_res = sm_4.fit_resample(X, y)
print('kmeans-smote:\t', Counter(y_res), '\n\n\n\n')
y_res = pd.DataFrame(y_res)
X_new = X_res.iloc[len(X):, :]
y_new = y_res.iloc[len(y):, :]
axes[2][1].scatter(X[0], X[1], c=y, alpha=0.5)
axes[2][1].scatter(X_new[0], X_new[1], c='red', alpha=0.2)
axes[2][1].set_title('kmeans_smote')
'''weight-svm-smote'''
sm_7 = all_smote_v2.SVMSMOTE(random_state=42, )
X_res, y_res = sm_7.fit_resample(X, y)
print('weight-SVM_smote:\t', Counter(y_res), '\n\n\n\n')
y_res = pd.DataFrame(y_res)
X_new = X_res.iloc[len(X):, :]
y_new = y_res.iloc[len(y):, :]
axes[3][0].scatter(X[0], X[1], c=y, alpha=0.5)
axes[3][0].scatter(X_new[0], X_new[1], c='red', alpha=0.2)
axes[3][0].set_title('weight_svm_smote')
'''SVM_SMOTE'''
sm_6 = over_sampling.SVMSMOTE(random_state=42, )
X_res, y_res = sm_6.fit_resample(X, y)
print('SVM_smote:\t', Counter(y_res), '\n\n\n\n')
y_res = pd.DataFrame(y_res)
X_new = X_res.iloc[len(X):, :]
y_new = y_res.iloc[len(y):, :]
axes[3][1].scatter(X[0], X[1], c=y, alpha=0.5)
axes[3][1].scatter(X_new[0], X_new[1], c='red', alpha=0.2)
axes[3][1].set_title('svm_smote')