def test_freqfeature_handles_nans_correctly(categorical_na):
freq_feature = FreqFeature()
result = freq_feature.fit_transform(categorical_na)
assert isinstance(result, pd.DataFrame)
assert len(categorical_na) == len(result)
assert set(categorical_na.columns) == set(result)
assert 0 == result.isna().sum().sum()
assert 1 / (len(categorical_na) - 1) == result.iloc[0, 0]
def test_freq_features_returns_0_when_unseen_value_is_given(categorical):
freq_feature = FreqFeature()
freq_feature.fit(categorical)
new_data = pd.DataFrame({"category_a": ["a1", "a2", "c25"]})
result = freq_feature.transform(new_data)
assert isinstance(result, pd.DataFrame)
assert len(new_data) == len(result)
assert 0 == result.iloc[-1, 0]
def test_freqfeature_returns_correctly(categorical):
freq_feature = FreqFeature()
result = freq_feature.fit_transform(categorical)
assert isinstance(result, pd.DataFrame)
assert len(categorical) == len(result)
assert set(categorical.columns) == set(result.columns)
for col in result.columns:
assert pd.api.types.is_numeric_dtype(result[col])
assert 1 / len(categorical) == result.iloc[0, 0]
assert all(1 == result.sum())
def test_freq_feature_can_be_used_in_cross_validation_string_data(
self, categorical: pd.DataFrame):
pipe = create_pipeline(FreqFeature())
score = cross_val_score(pipe,
categorical,
np.array([1, 0, 1, 0]),
cv=2)
assert np.all(score >= 0)
def test_freq_feature_can_be_used_in_grid_search(
self, categorical: pd.DataFrame):
pipe = create_pipeline(FreqFeature())
model = GridSearchCV(
pipe,
param_grid={"clf__strategy": ["stratified", "most_frequent"]},
cv=2)
model.fit(categorical, [1, 0, 1, 0])
assert hasattr(model, "best_estimator_")
def test_works_without_args(self):
assert FreqFeature()