Develop

multi_imbalance/datasets/_data_loader.py

@@ -24,7 +24,7 @@
 
 import numpy as np
 
-from sklearn.datasets.base import Bunch
+from sklearn.datasets._base import Bunch
 
 PRE_FILENAME = 'x'
 POST_FILENAME = 'data.npz'

multi_imbalance/ensemble/tests/test_mrbbagging.py

@@ -1,7 +1,7 @@
 import unittest
 from unittest.mock import MagicMock
 
-from sklearn.tree import tree
+from sklearn.tree import DecisionTreeClassifier
 
 from multi_imbalance.ensemble.mrbbagging import MRBBagging
 import numpy as np
@@ -36,47 +36,47 @@
 
 class TestMRBBagging(unittest.TestCase):
  def test_api(self):
- mrbbagging = MRBBagging(1, tree.DecisionTreeClassifier(random_state=0), random_state=0)
+ mrbbagging = MRBBagging(1, DecisionTreeClassifier(random_state=0), random_state=0)
  mrbbagging.fit(X_train, y_train)
  y_pred = mrbbagging.predict(X_test)
  assert all(y_pred == y_test)
 
  def test_api_multiple_trees(self):
- mrbbagging = MRBBagging(5, tree.DecisionTreeClassifier(random_state=0), random_state=0)
+ mrbbagging = MRBBagging(5, DecisionTreeClassifier(random_state=0), random_state=0)
  mrbbagging.fit(X_train, y_train)
  y_pred = mrbbagging.predict(X_test)
  assert all(y_pred == y_test)
 
  def test_api_with_feature_selection(self):
- mrbbagging = MRBBagging(1, tree.DecisionTreeClassifier(random_state=0), feature_selection=True, random_state=0)
+ mrbbagging = MRBBagging(1, DecisionTreeClassifier(random_state=0), feature_selection=True, random_state=0)
  mrbbagging.fit(X_train, y_train)
  y_pred = mrbbagging.predict(X_test)
  assert all(y_pred == y_test)
 
  def test_api_with_random_feature_selection(self):
- mrbbagging = MRBBagging(1, tree.DecisionTreeClassifier(random_state=0), feature_selection=True, random_fs=True,
+ mrbbagging = MRBBagging(1, DecisionTreeClassifier(random_state=0), feature_selection=True, random_fs=True,
  random_state=0)
  mrbbagging.fit(X_train, y_train)
  y_pred = mrbbagging.predict(X_test)
  assert all(y_pred == y_test)
 
  def test_api_with_feature_selection_sqrt_features(self):
- mrbbagging = MRBBagging(1, tree.DecisionTreeClassifier(random_state=0), feature_selection=True,
+ mrbbagging = MRBBagging(1, DecisionTreeClassifier(random_state=0), feature_selection=True,
  half_features=False, random_state=0)
  mrbbagging.fit(X_train, y_train)
  y_pred = mrbbagging.predict(X_test)
  assert all(y_pred == y_test)
 
  def test__group_data(self):
- mrbbagging = MRBBagging(1, tree.DecisionTreeClassifier())
+ mrbbagging = MRBBagging(1, DecisionTreeClassifier())
  x = [[1, 1, 1], [2, 2, 2], [3, 3, 3]]
  y = ["A", "B", "C"]
  classes, grouped_data = mrbbagging._group_data(x, y)
  self.assertEqual(classes, {'A', 'B', 'C'})
  self.assertEqual(grouped_data, {'C': [[[3, 3, 3], 'C']], 'A': [[[1, 1, 1], 'A']], 'B': [[[2, 2, 2], 'B']]})
 
  def test__group_data_with_none(self):
- mrbbagging = MRBBagging(1, tree.DecisionTreeClassifier())
+ mrbbagging = MRBBagging(1, DecisionTreeClassifier())
  x = [[1, 1, 1], [2, 2, 2], [3, 3, 3]]
  y = ["A", None, "C"]
  with self.assertRaises(AssertionError):
@@ -95,7 +95,7 @@ def test_fit_with_invalid_classifier(self):
 
  def test_with_invalid_k(self):
  with self.assertRaises(AssertionError):
- MRBBagging(0, tree.DecisionTreeClassifier())
+ MRBBagging(0, DecisionTreeClassifier())
 
 
 if __name__ == '__main__':

multi_imbalance/resampling/static_smote.py

@@ -2,9 +2,10 @@
 
 import numpy as np
 from imblearn.over_sampling import SMOTE
+from imblearn.base import BaseSampler
 
 
-class StaticSMOTE:
+class StaticSMOTE(BaseSampler):
  """
  Static SMOTE implementation:
 
@@ -13,9 +14,21 @@ class StaticSMOTE:
  procedure based on sensitivity for multi-class problems. Pattern Recognit. 44, 1821–1833
  (2011)
  """
+ def __init__(self):
+ super().__init__()
+ self._sampling_type = 'over-sampling'
 
- # TODO add docstring
- def fit_transform(self, X, y):
+ def _fit_resample(self, X, y):
+ """
+ Performs resampling
+
+ :param X:
+ two dimensional numpy array (number of samples x number of features) with float numbers
+ :param y:
+ one dimensional numpy array with labels for rows in X
+ :return:
+ Resampled X and y as numpy arrays
+ """
  cnt = Counter(y)
  min_class = min(cnt, key=cnt.get)
  X_original, y_original = X.copy(), y.copy()

multi_imbalance/resampling/tests/test_static_smote.py

@@ -12,7 +12,7 @@ def test_static_smote():
 
  y = np.array([1] * 100 + [2] * 30 + [3] * 20)
  ssm = StaticSMOTE()
- X_resampled, y_resampled = ssm.fit_transform(X, y)
+ X_resampled, y_resampled = ssm.fit_resample(X, y)
  cnt = Counter(y_resampled)
  assert cnt[1] == 100
  assert cnt[2] == 60

setup.py

@@ -26,7 +26,7 @@
  ],
  install_requires=[
  "numpy>=1.17.0",
- "scikit-learn==0.21.3",
+ "scikit-learn>=0.22.1",
  "pandas>=0.25.1",
  "pytest>=5.1.2",
  "imbalanced-learn>=0.6.1",