def test_triple_add(self):
X = pd.DataFrame({'a': [1, 2, 3], 'b': [2, 3, 4]})
y = pd.Series([1, 2, 3])
prd = pd_preprocessing.MinMaxScaler() | \
trans() + trans() + trans() | \
pd_linear_model.LinearRegression()
y_hat = prd.fit(X, y).predict(X)
self.assertTrue(isinstance(y_hat, pd.Series))
prd = pd_preprocessing.MinMaxScaler() | \
trans() + (trans() + trans()) | \
pd_linear_model.LinearRegression()
y_hat = prd.fit(X, y).predict(X)
self.assertTrue(isinstance(y_hat, pd.Series))
def test_pipe_fit(self):
X = pd.DataFrame({'a': [1, 2, 3]})
y = pd.Series([1, 2, 3])
prd = pd_preprocessing.StandardScaler() | \
pd_preprocessing.StandardScaler() | \
pd_linear_model.LinearRegression()
y_hat = prd.fit(X, y).predict(X)
self.assertTrue(isinstance(y_hat, pd.Series))
prd = pd_preprocessing.StandardScaler() | \
(pd_preprocessing.StandardScaler() | \
pd_linear_model.LinearRegression())
y_hat = prd.fit(X, y).predict(X)
self.assertTrue(isinstance(y_hat, pd.Series))
def test_neut(self):
prd = linear_model.LinearRegression(fit_intercept=False)
self.assertIn('get_params', dir(prd))
self.assertEqual(prd.get_params()['fit_intercept'], False)
prd = pd_linear_model.LinearRegression(fit_intercept=False)
self.assertIn('get_params', dir(prd))
self.assertEqual(prd.get_params()['fit_intercept'], False)
def test_getattr(self):
x = pd.DataFrame({'a': [1, 2, 3]})
y = pd.Series([1, 2, 3])
prd = pd_linear_model.LinearRegression()
with self.assertRaises(AttributeError):
prd.coef_
prd.fit(x, y)
prd.coef_
def test_pipeline_fit_internal_pd_stage(self):
X = pd.DataFrame({'a': [1, 2, 3]})
y = pd.Series([1, 2, 3])
p = pd_pipeline.make_pipeline(pd_linear_model.LinearRegression())
self.assertTrue(isinstance(p, FrameMixin))
pd_p = frame(p)
y_hat = pd_p.fit(X, y).predict(X)
self.assertTrue(isinstance(y_hat, pd.Series))
def test_fit_predict(self):
X = pd.DataFrame({'a': [1, 2, 3]})
y = pd.Series([1, 2, 3])
pd_y_hat = pd_linear_model.LinearRegression().fit(X, y).predict(X)
self.assertTrue(isinstance(pd_y_hat, pd.Series))
y_hat = linear_model.LinearRegression().fit(X, y).predict(X)
self.assertFalse(isinstance(y_hat, pd.Series))
np.testing.assert_equal(pd_y_hat, y_hat)
def test_fit_permute_cols(self):
X = pd.DataFrame({'a': [1, 2, 3], 'b': [30, 23, 2]})
y = pd.Series([1, 2, 3])
pred = pd_linear_model.LinearRegression().fit(X, y)
pd_y_hat = pred.predict(X[['b', 'a']])
self.assertTrue(isinstance(pd_y_hat, pd.Series))
y_hat = linear_model.LinearRegression().fit(X, y).predict(X)
self.assertFalse(isinstance(y_hat, pd.Series))
np.testing.assert_equal(pd_y_hat, y_hat)
def test_pipeline_fit(self):
X = pd.DataFrame({'a': [1, 2, 3]})
y = pd.Series([1, 2, 3])
# Tmp Ami - make verify that are framemixins
p = pd_pipeline.make_pipeline(pd_linear_model.LinearRegression())
self.assertTrue(isinstance(p, FrameMixin))
pd_p = frame(p)
pd_p = pd_p.fit(X, y)
y_hat = pd_p.fit(X, y).predict(X)
self.assertTrue(isinstance(y_hat, pd.Series))
def test_fit_bad_cols(self):
X = pd.DataFrame({'a': [1, 2, 3], 'b': [30, 23, 2]})
y = pd.Series([1, 2, 3])
pred = pd_linear_model.LinearRegression().fit(X, y)
y_hat = pred.predict(X)
self.assertTrue(isinstance(y_hat, pd.Series))
X.rename(columns={'a': 'c'}, inplace=True)
with self.assertRaises(KeyError):
pred.predict(X)
def test_cross_val_predict(self):
from ibex.sklearn import model_selection as pd_model_selection
iris, features = _load_iris()
n = 100
df = pd.DataFrame({
'x': range(n),
'y': range(n),
},
index=['i%d' % i for i in range(n)])
y_hat = pd_model_selection.cross_val_predict(
pd_linear_model.LinearRegression(), df[['x']], df['y'])
self.assertIsInstance(y_hat, pd.Series)
self.assertEqual(len(y_hat), len(df))
def test_direct_pipe_adapter(self):
clf = pd_decomposition.PCA() | pd_linear_model.LinearRegression()
unpickled_clf = pickle.loads(pickle.dumps(clf))
def __init__(self):
self.prd = pd_linear_model.LinearRegression()
_Xs.append(_iris[_features])
_ys.append(_iris['class'])
_Xs.append(_iris[_features])
_ys.append(_iris['class'] == _iris['class'].values[0])
_iris = _iris.copy()
_iris.index = ['i%d' % i for i in range(len(_iris))]
_dataset_names.append('iris_str_index')
_Xs.append(_iris[_features])
_ys.append(_iris['class'])
_estimators = []
_estimators.append(
(preprocessing.StandardScaler(), pd_preprocessing.StandardScaler(), True))
_estimators.append((decomposition.PCA(), pd_decomposition.PCA(), True))
_estimators.append((linear_model.LinearRegression(),
frame(pd_linear_model.LinearRegression()), True))
_estimators.append((linear_model.LinearRegression(),
pd_linear_model.LinearRegression(), True))
_estimators.append(
(pipeline.make_pipeline(decomposition.PCA(),
linear_model.LinearRegression()),
pd_decomposition.PCA() | pd_linear_model.LinearRegression(), True))
_estimators.append(
(pipeline.make_pipeline(feature_selection.SelectKBest(k=2),
decomposition.PCA(),
linear_model.LinearRegression()),
pd_feature_selection.SelectKBest(k=2) | pd_decomposition.PCA()
| pd_linear_model.LinearRegression(), True))
_estimators.append(
(pipeline.make_pipeline(feature_selection.SelectKBest(k=2),
decomposition.PCA(),
def setUpClass(cls):
cls._prd = pd_linear_model.LinearRegression()
cls._clf = pd_linear_model.LogisticRegression()
def test_make_pipeline(self):
p = pd_pipeline.make_pipeline(pd_preprocessing.StandardScaler(),
pd_linear_model.LinearRegression())
