def test_features_raw_data_no_target(data_classification_balanced, feature_descriptor):
"""Testing if raw_dataframe() drops Target column when drop_target=True."""
X, y = data_classification_balanced
f = Features(X, y, feature_descriptor)
expected_df = X.drop(["Date"], axis=1)
cols = expected_df.columns
actual_df = f.raw_data(drop_target=True)[cols]
assert actual_df.equals(expected_df)
def test_features_create_mapped_dataframe(data_classification_balanced, feature_descriptor, expected_raw_mapping):
"""Testing if ._create_mapped_dataframe correctly returns mapped dataframe (with replaced values according to
mapping). """
X, y = data_classification_balanced
f = Features(X, y, feature_descriptor)
expected_df = pd.concat([X, y], axis=1).drop(["Date"], axis=1).replace(expected_raw_mapping)
cols = expected_df.columns
actual_df = f._create_mapped_dataframe()[cols]
assert actual_df.equals(expected_df)
def test_features_numerical_features_no_target(
feature_list, target, expected, data_classification_balanced, feature_descriptor
):
"""Testing if .numerical_features() returns correct values when drop_target = True (without Target feature name)."""
X, y = data_classification_balanced
f = Features(X, y, feature_descriptor)
f._numerical_features = feature_list
f.target = target
actual = f.numerical_features(drop_target=True)
assert actual == expected
def test_features_data(data_classification_balanced, feature_descriptor, expected_raw_mapping):
"""Testing if .data() returns mapped df (with replaced values according to mapping) but without Target column (
when drop_target=True). """
X, y = data_classification_balanced
f = Features(X, y, feature_descriptor)
expected_df = X.drop(["Date"], axis=1).replace(expected_raw_mapping)
cols = expected_df.columns
actual_df = f.data(drop_target=True)[cols]
assert actual_df.equals(expected_df)
def test_features_create_mapping(data_classification_balanced, feature_descriptor, expected_mapping):
"""Testing if ._create_mapping() creates a correct mapping dictionary."""
X, y = data_classification_balanced
f = Features(X, y, feature_descriptor)
expected = expected_mapping
for feat in ["Height", "Price"]:
expected[feat] = None
actual = f.mapping()
assert actual == expected
def test_features_create_raw_dataframe(data_classification_balanced, feature_descriptor):
"""Testing if .create_raw_dataframe returns correct dataframe (the same that was provided as input to the
object). """
X, y = data_classification_balanced
f = Features(X, y, feature_descriptor)
expected_df = pd.concat([X, y], axis=1).drop(["Date"], axis=1)
cols = expected_df.columns
actual_df = f._create_raw_dataframe()[cols]
assert actual_df.equals(expected_df)
def test_features_raw_data_excluded_transformed(data_classification_balanced, feature_descriptor, transformed_columns):
"""Testing if raw_data returns correct dataframe without transformed columns when excluded_transformed
is set to True."""
X, y = data_classification_balanced
f = Features(X, y, feature_descriptor, transformed_columns)
expected_df = X.drop(["Date"], axis=1)
if transformed_columns:
expected_df = expected_df.drop(transformed_columns, axis=1)
cols = expected_df.columns
actual_df = f.raw_data(exclude_transformed=True)[cols]
assert actual_df.equals(expected_df)
def test_features_numerical_features_exclude_transformed(
data_classification_balanced, feature_descriptor, transformed_features
):
"""Testing if returning numerical features list with transformed columns excluded works properly."""
col_list = ["Height", "Price"]
X, y = data_classification_balanced
f = Features(X, y, feature_descriptor, transformed_features)
actual_result = f.numerical_features(exclude_transformed=True)
if transformed_features:
expected_result = [feature for feature in col_list if feature not in transformed_features]
else:
expected_result = col_list
assert actual_result == expected_result
def test_features_categorical_features_exclude_transformed(
data_classification_balanced, feature_descriptor, transformed_features
):
"""Testing if returning categorical features list with transformed columns excluded works properly."""
col_list = ["AgeGroup", "bool", "Product", "Sex", "Target"]
X, y = data_classification_balanced
f = Features(X, y, feature_descriptor, transformed_features)
actual_result = f.categorical_features(exclude_transformed=True)
if transformed_features:
expected_result = [feature for feature in col_list if feature not in transformed_features]
else:
expected_result = col_list
assert actual_result == expected_result
def test_features_analyze_features_transformed_features(
data_classification_balanced, feature_descriptor, transformed_features
):
"""Testing if creating features properly assigns Transformed flag based on provided transformed_features
sequence."""
X, y = data_classification_balanced
f = Features(X, y, feature_descriptor, transformed_features)
f.original_dataframe = pd.concat([X, y], axis=1) # original_dataframe needs to be set up
actual = f._analyze_features(feature_descriptor)
for feature in actual.keys():
if feature in transformed_features:
assert actual[feature].transformed
else:
assert not actual[feature].transformed
def test_features_create_descriptions(data_classification_balanced, feature_descriptions, feature_descriptor):
"""Testing if ._create_descriptions creates a correct descriptions dictionary."""
placeholder = Features._description_not_available
d = FeatureDescriptor._description
assert True
expected_descriptions = {}
for feat in ["Sex", "Height", "Product", "Price", "bool", "Target"]:
expected_descriptions[feat] = feature_descriptions[feat][d]
expected_descriptions["AgeGroup"] = placeholder
X, y = data_classification_balanced
f = Features(X, y, feature_descriptor)
actual_descriptions = f.descriptions()
assert actual_descriptions == expected_descriptions
def test_features_create_raw_dataframe_preserving_index(data_classification_balanced, feature_descriptor):
"""Testing if create_raw_dataframe preserves the index of the DataFrame."""
X, y = data_classification_balanced
not_expected_df = pd.concat([X, y], axis=1).drop(["Date"], axis=1)
length = X.shape[0]
new_ind = list(range(100, length + 100))
X.index = new_ind
y.index = new_ind
f = Features(X, y, feature_descriptor)
expected_df = pd.concat([X, y], axis=1).drop(["Date"], axis=1)
expected_df.index = new_ind
cols = expected_df.columns
actual_df = f._create_raw_dataframe()[cols]
assert not actual_df.equals(not_expected_df)
assert actual_df.equals(expected_df)
def test_features_features_list_no_target(
data_classification_balanced, feature_descriptor_type, feature_descriptor, feature_descriptor_forced_categories
):
"""Testing if .features() returns correct values when drop_target = True (without Target feature name)."""
expected = ["AgeGroup", "bool", "Height", "Price", "Product", "Sex"]
X, y = data_classification_balanced
# couldn't find a way to incorporate fixtures into @pytest.mark.parametrize
if feature_descriptor_type == "normal":
fd = feature_descriptor
elif feature_descriptor_type == "forced":
fd = feature_descriptor_forced_categories
else:
raise
f = Features(X, y, fd)
actual = f.features(drop_target=True)
assert actual == expected
def test_features_create_categorical_features(
data_classification_balanced, feature_descriptor_type, expected,
feature_descriptor, feature_descriptor_forced_categories
):
"""Testing if ._create_categorical_features() returns correct values depending on the Features provided."""
X, y = data_classification_balanced
# couldn't find a way to incorporate fixtures into @pytest.mark.parametrize
if feature_descriptor_type == "normal":
fd = feature_descriptor
elif feature_descriptor_type == "forced":
fd = feature_descriptor_forced_categories
else:
raise
f = Features(X, y, fd)
actual = f._create_categorical_features()
assert actual == expected
def test_analyzer_numeric_describe_no_numerical_features(
data_classification_balanced):
"""Testing if numeric_describe() returns None when there are no numerical columns present."""
numerical_cols = ["Price", "Height"]
X, y = data_classification_balanced
X = X.drop(numerical_cols, axis=1)
f = Features(X, y, None)
analyzer = Analyzer(f)
actual_result = analyzer.numerical_describe_df()
assert actual_result is None
def test_features_impute_column_type(data_classification_balanced, column_name, expected_type):
"""Testing if imputing column type works correctly."""
X = data_classification_balanced[0]
y = data_classification_balanced[1]
df = pd.concat([X, y], axis=1)
f = Features(X, y)
cat = f._categorical
num = f._numerical
dat = f._date
if expected_type == "categorical":
expected = cat
elif expected_type == "numerical":
expected = num
elif expected_type == "date":
expected = dat
else:
raise
actual = f._impute_column_type(df[column_name])
assert actual == expected
def test_features_analyze_features(data_classification_balanced, feature_descriptor):
"""Testing if .analyze_features() method of Features class returns a dictionary with a correct content"""
n = NumericalFeature
c = CategoricalFeature
expected = {
"Sex": c,
"AgeGroup": c,
"Height": n,
"Product": c,
"Price": n,
"bool": c,
"Target": c
}
X, y = data_classification_balanced
f = Features(X, y, feature_descriptor)
f.original_dataframe = pd.concat([X, y], axis=1) # original_dataframe needs to be set up
actual = f._analyze_features(feature_descriptor)
assert isinstance(actual, dict)
for key, item in expected.items():
assert isinstance(actual[key], item)
def test_analyzer_categorical_describe_no_categorical_features(
data_classification_balanced):
"""Testing if categorical_describe() returns None when there are no categorical columns present."""
numerical_cols = [
"Price", "Height"
] # its easier to provide numerical columns instead of dropping all categorical
data = data_classification_balanced[0]
X = data[numerical_cols[:1]]
y = data[numerical_cols[1]]
f = Features(X, y, None)
analyzer = Analyzer(f)
actual_result = analyzer.categorical_describe_df()
assert actual_result is None
def fixture_features(data_classification_balanced, feature_descriptor):
"""Fixture Features object for data_classification_balanced test data."""
X, y = data_classification_balanced
f = Features(X, y, feature_descriptor)
return f
def fixture_features_multiclass(data_multiclass, feature_descriptor):
"""Fixture Features for multiclass problem."""
X, y = data_multiclass
f = Features(X, y, feature_descriptor)
return f
def test_features_unused_features(data_classification_balanced, feature_descriptor):
"""Testing if unused_features() returns correct values."""
X, y = data_classification_balanced
f = Features(X, y, feature_descriptor)
assert f.unused_features() == ["Date"]