def test_pipeline_savefig_poof(self):
"""
Test the poof call with an outdir to save all the figures
"""
pipeline = VisualPipeline(
[
("a", mock.MagicMock(MockTransformer())),
("b", mock.MagicMock(MockVisualTransformer())),
("c", mock.MagicMock(MockTransformer())),
("d", mock.MagicMock(MockVisualTransformer())),
("e", mock.MagicMock(MockVisualEstimator())),
]
)
# Must use path joining for Windows compatibility
tmpdir = os.path.join("tmp", "figures")
pipeline.poof(outdir=tmpdir)
pipeline.steps[1][1].poof.assert_called_once_with(
outpath=os.path.join(tmpdir, "b.pdf")
)
pipeline.steps[3][1].poof.assert_called_once_with(
outpath=os.path.join(tmpdir, "d.pdf")
)
pipeline.steps[4][1].poof.assert_called_once_with(
outpath=os.path.join(tmpdir, "e.pdf")
)
def test_fit_transform_poof_and_draw_calls(self):
"""
Test calling fit, transform, draw and poof on the pipeline
"""
pipeline = VisualPipeline([
('a', MockTransformer()),
('b', MockVisualTransformer()),
('c', MockTransformer()),
('d', MockVisualTransformer()),
('e', MockEstimator()),
])
X = [[1, 1, 1, 1, 1], [2, 2, 2, 2, 2], [3, 3, 3, 3, 3]]
y = [1, 2, 3, 4, 5]
pipeline.fit(X, y)
for name, step in pipeline.named_steps.items():
step.fit.assert_called_once_with(X, y)
pipeline.transform(X)
for name, step in pipeline.named_steps.items():
step.transform.assert_called_once_with(X)
pipeline.draw()
for name, step in pipeline.named_steps.items():
if name in {'a', 'c', 'e'}: continue
step.draw.assert_called_once_with()
pipeline.poof()
for name, step in pipeline.named_steps.items():
if name in {'a', 'c', 'e'}: continue
step.poof.assert_called_once_with()
def test_validate_steps(self):
"""
Assert that visual transformers can be added to pipelines
"""
# Pipeline objects have a _validate_steps method that raises an
# TypeError if the steps don't match transforms --> estimator.
# validate a bad intermediate transformer on the Pipeline
with pytest.raises(TypeError):
Pipeline(
[
("real", MockTransformer()),
("bad", Thing()),
("model", MockEstimator()),
]
)
# validate a bad intermediate transformer on the VisualPipeline
with pytest.raises(TypeError):
VisualPipeline(
[
("real", MockTransformer()),
("bad", Thing()),
("model", MockEstimator()),
]
)
# validate a bad final estimator on the Pipeline
with pytest.raises(TypeError):
Pipeline([("real", MockTransformer()), ("bad", Thing())])
# validate a bad final estimator on the VisualPipeline
with pytest.raises(TypeError):
VisualPipeline([("real", MockTransformer()), ("bad", Thing())])
# validate visual transformers on a Pipeline
try:
Pipeline(
[
("real", MockTransformer()),
("visual", MockVisualTransformer()),
("model", MockEstimator()),
]
)
except TypeError:
self.fail("could not add a visual transformer to a Pipeline!")
# validate visual transformers on a VisualPipeline
try:
VisualPipeline(
[
("real", MockTransformer()),
("visual", MockVisualTransformer()),
("model", MockEstimator()),
]
)
except TypeError:
self.fail("could not add a visual transformer to a VisualPipeline!")
def test_fit_transform_show_and_draw_calls(self):
"""
Test calling fit, transform, and show on the pipeline
"""
pipeline = VisualPipeline([
("a", mock.MagicMock(MockTransformer())),
("b", mock.MagicMock(MockVisualTransformer())),
("c", mock.MagicMock(MockTransformer())),
("d", mock.MagicMock(MockVisualTransformer())),
("e", mock.MagicMock(MockEstimator())),
])
X = [[1, 1, 1, 1, 1], [2, 2, 2, 2, 2], [3, 3, 3, 3, 3]]
y = [1, 2, 3, 4, 5]
pipeline.fit(X, y)
for name, step in pipeline.named_steps.items():
step.fit.assert_called_once_with(X, y)
pipeline.transform(X)
for name, step in pipeline.named_steps.items():
if name == "e":
continue
step.transform.assert_called_once_with(X)
pipeline.show()
for name, step in pipeline.named_steps.items():
if name in {"a", "c", "e"}:
continue
step.show.assert_called_once_with(outpath=None)
def test_pipeline_show(self):
"""
Test the show call against the VisualPipeline
"""
pipeline = VisualPipeline([
("a", mock.MagicMock(MockTransformer())),
("b", mock.MagicMock(MockVisualTransformer())),
("c", mock.MagicMock(MockTransformer())),
("d", mock.MagicMock(MockVisualTransformer())),
("e", mock.MagicMock(MockEstimator())),
])
pipeline.show()
pipeline.steps[1][1].show.assert_called_once_with(outpath=None)
pipeline.steps[3][1].show.assert_called_once_with(outpath=None)
def test_pipeline_savefig_poof(self):
"""
Test the poof call with an outdir to save all the figures
"""
pipeline = VisualPipeline([
('a', mock.MagicMock(MockTransformer())),
('b', mock.MagicMock(MockVisualTransformer())),
('c', mock.MagicMock(MockTransformer())),
('d', mock.MagicMock(MockVisualTransformer())),
('e', mock.MagicMock(MockVisualEstimator()),)
])
pipeline.poof(outdir="/tmp/figures")
pipeline.steps[1][1].poof.assert_called_once_with(outpath="/tmp/figures/b.pdf")
pipeline.steps[3][1].poof.assert_called_once_with(outpath="/tmp/figures/d.pdf")
pipeline.steps[4][1].poof.assert_called_once_with(outpath="/tmp/figures/e.pdf")
def test_pipeline_poof(self):
"""
Test the poof call against the VisualPipeline
"""
pipeline = VisualPipeline([
('a', mock.MagicMock(MockTransformer())),
('b', mock.MagicMock(MockVisualTransformer())),
('c', mock.MagicMock(MockTransformer())),
('d', mock.MagicMock(MockVisualTransformer())),
('e', mock.MagicMock(MockEstimator()),)
])
pipeline.poof()
pipeline.steps[1][1].poof.assert_called_once_with(outpath=None)
pipeline.steps[3][1].poof.assert_called_once_with(outpath=None)
def test_pipeline_poof(self):
"""
Test the poof call against the VisualPipeline
"""
pipeline = VisualPipeline([
('a', mock.MagicMock(MockTransformer())),
('b', mock.MagicMock(MockVisualTransformer())),
('c', mock.MagicMock(MockTransformer())),
('d', mock.MagicMock(MockVisualTransformer())),
('e', mock.MagicMock(MockEstimator()),)
])
pipeline.poof()
pipeline.steps[1][1].poof.assert_called_once_with(outpath=None)
pipeline.steps[3][1].poof.assert_called_once_with(outpath=None)
def test_fit_transform_poof_and_draw_calls(self):
"""
Test calling fit, transform, and poof on the pipeline
"""
pipeline = VisualPipeline([
('a', mock.MagicMock(MockTransformer())),
('b', mock.MagicMock(MockVisualTransformer())),
('c', mock.MagicMock(MockTransformer())),
('d', mock.MagicMock(MockVisualTransformer())),
('e', mock.MagicMock(MockEstimator()),)
])
X = [[1, 1, 1, 1, 1],
[2, 2, 2, 2, 2],
[3, 3, 3, 3, 3]]
y = [1, 2, 3, 4, 5]
pipeline.fit(X, y)
for name, step in pipeline.named_steps.items():
step.fit.assert_called_once_with(X, y)
pipeline.transform(X)
for name, step in pipeline.named_steps.items():
if name == 'e': continue
step.transform.assert_called_once_with(X)
pipeline.poof()
for name, step in pipeline.named_steps.items():
if name in {'a', 'c', 'e'}: continue
step.poof.assert_called_once_with(outpath=None)
def test_pipeline_savefig_poof(self):
"""
Test the poof call with an outdir to save all the figures
"""
pipeline = VisualPipeline([
('a', mock.MagicMock(MockTransformer())),
('b', mock.MagicMock(MockVisualTransformer())),
('c', mock.MagicMock(MockTransformer())),
('d', mock.MagicMock(MockVisualTransformer())),
('e', mock.MagicMock(MockVisualEstimator()),)
])
# Must use path joining for Windows compatibility
tmpdir = os.path.join("tmp", "figures")
pipeline.poof(outdir=tmpdir)
pipeline.steps[1][1].poof.assert_called_once_with(outpath=os.path.join(tmpdir, "b.pdf"))
pipeline.steps[3][1].poof.assert_called_once_with(outpath=os.path.join(tmpdir, "d.pdf"))
pipeline.steps[4][1].poof.assert_called_once_with(outpath=os.path.join(tmpdir, "e.pdf"))
def selectDiscr():
data_path = "labeled_data.csv"
data = pd.read_csv(data_path)
# We create the preprocessing pipelines for both numeric and categorical data.
numeric_features = ['count_reviews', 'rating']
numeric_transformer = Pipeline(
steps=[('imputer',
SimpleImputer(strategy='median')), ('scaler',
StandardScaler())])
categorical_features = ['product_category']
categorical_transformer = Pipeline(
steps=[('imputer',
SimpleImputer(strategy='constant', fill_value='missing')
), ('onehot', OneHotEncoder(handle_unknown='ignore'))])
preprocessor = ColumnTransformer(
transformers=[('num', numeric_transformer, numeric_features
), ('cat', categorical_transformer,
categorical_features)])
viz = DiscriminationThreshold(LogisticRegression())
clf = VisualPipeline(steps=[
('preprocessor', preprocessor),
#('classifier', LogisticRegression(solver='lbfgs')),
('viz', viz)
])
X = data.drop('label', axis=1)
y = data['label']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
model = clf.fit(X_train, y_train)
model.poof()
def test_visual_steps_property(self):
"""
Test the visual steps property to filter visualizers
"""
pipeline = VisualPipeline([
('a', MockTransformer()),
('b', VisualTransformerSpec()),
('c', MockTransformer()),
('d', VisualTransformerSpec()),
('e', MockEstimator()),
])
self.assertIn('b', pipeline.visual_steps)
self.assertIn('d', pipeline.visual_steps)
def test_visual_steps_property(self):
"""
Test the visual steps property to filter visualizers
"""
pipeline = VisualPipeline([
("a", MockTransformer()),
("b", MockVisualTransformer()),
("c", MockTransformer()),
("d", MockVisualTransformer()),
("e", MockEstimator()),
])
assert "a" not in pipeline.visual_steps
assert "b" in pipeline.visual_steps
assert "c" not in pipeline.visual_steps
assert "d" in pipeline.visual_steps
assert "e" not in pipeline.visual_steps
def modelSelection():
data_path = "labeled_data.csv"
data = pd.read_csv(data_path)
# Preprocessing pipelines for both numeric and categorical data.
numeric_features = ['count_reviews', 'rating']
numeric_transformer = Pipeline(
steps=[('imputer',
SimpleImputer(strategy='median')), ('scaler',
StandardScaler())])
categorical_features = ['product_category']
categorical_transformer = Pipeline(
steps=[('imputer',
SimpleImputer(strategy='constant', fill_value='missing')
), ('onehot', OneHotEncoder(handle_unknown='ignore'))])
preprocessor = ColumnTransformer(
transformers=[('num', numeric_transformer, numeric_features
), ('cat', categorical_transformer,
categorical_features)])
X = data.drop('label', axis=1)
y = data['label']
fig = plt.figure()
ax = fig.add_subplot()
#viz_ridge = ClassificationReport(RidgeClassifier(), classes = ['not recommended', 'recommended'], support=True)
viz_logistic = ClassificationReport(
LogisticRegression(),
#SGDClassifier(),
#RidgeClassifier(),
classes=['not recommended', 'recommended'],
support=True)
#clf_ridge = VisualPipeline(steps=[('preprocessor', preprocessor),
# #('classifier', LogisticRegression(solver='lbfgs')),
# ('viz', viz_ridge)])
#Visual Pipeline is used to visualize the report
clf_logistic = VisualPipeline(steps=[
('preprocessor', preprocessor),
#('classifier', LogisticRegression(solver='lbfgs')),
('viz', viz_logistic)
])
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
#model_ridge = clf_ridge.fit(X_train, y_train)
model_logistic = clf_logistic.fit(X_train, y_train)
#preds_ridge = clf_ridge.predict(X_test)
preds_logistic = clf_logistic.predict(X_test)
#print("RidgeClassifier model score: %.3f" % clf_ridge.score(X_test, y_test))
print("LogisticRegression model score: %.3f" %
clf_logistic.score(X_test, y_test))
#clf_ridge.poof()
clf_logistic.poof()
# Evaluate accuracy
#print("RidgeClassifier accuracy: ", accuracy_score(y_test, preds_ridge))
print("LogisticRegression accuracy: ",
accuracy_score(y_test, preds_logistic))
final_predictions = X_test
final_predictions['target'] = y_test
final_predictions['prediction'] = preds_logistic
#print(final_predictions)
filename = 'model_products.sav'
joblib.dump(model_logistic, filename)