def test_transform(self):
# When there is no classifier assigned
tx = get_transformer('hard')
with self.assertRaises(AssertionError):
tx.transform(self.X)
# When there is a classifier assigned
clf = RandomForestClassifier(n_estimators=10)
clf.fit(self.X, self.Y)
tx = get_transformer('hard', clf)
transformed = tx.transform(self.X)
self.assertIsInstance(transformed, numpy.ndarray)
self.assertEqual(transformed.shape[1], 1)
self.assertEqual(transformed.shape[0], self.X.shape[0])
self.assertEqual(2, len(transformed.shape))
def compile(self, n_jobs=2) -> Pipeline:
"""
To Build the ensemble. This method should be called only if all models in base
are already fitted.
:return:
"""
meta_features = []
for i, mdl in enumerate(self.models):
meta_features.append((mdl.__class__.__name__ + str(i),
get_transformer(kind=self._get_voting(i),
clf=mdl)))
self.ensemble = Pipeline(
steps=[('base_layer', FeatureUnion(meta_features, n_jobs=n_jobs)
), ('final_layer', self.meta_model)])
return self.ensemble
def test_fit(self):
clf = RandomForestClassifier(n_estimators=10)
clf.fit(self.X, self.Y)
tx = get_transformer('soft', clf)
fit = tx.fit(self.X, self.Y)
self.assertIsInstance(fit, SoftClassifierTransformer)
def test_get(self):
tx = get_transformer('soft')
self.assertIsInstance(tx, SoftClassifierTransformer)
def test_get(self):
tx = get_transformer('hard')
self.assertIsInstance(tx, HardClassifierTransformer)