def test_imputing_all_missing_values_categorical(missing):
"""Assert that all missing values are imputed in categorical columns."""
X = [[missing, "a", "a"], ["b", "c", missing], ["b", "a", "c"],
["c", "a", "a"]]
y = [1, 1, 0, 0]
imputer = Imputer(strat_cat="most_frequent")
X, y = imputer.fit_transform(X, y)
assert X.isna().sum().sum() == 0
def test_imputing_all_missing_values_numeric(missing):
"""Assert that all missing values are imputed in numeric columns."""
X = [[missing, 1, 1], [2, 5, 2], [4, missing, 1], [2, 1, 1]]
y = [1, 1, 0, 0]
imputer = Imputer(strat_num="mean")
imputer.missing.append(99)
X, y = imputer.fit_transform(X, y)
assert X.isna().sum().sum() == 0
def test_cols_too_many_nans():
"""Assert that columns with too many missing values are dropped."""
X = X_bin.copy()
for i in range(5): # Add 5 cols with all NaN values
X["col " + str(i)] = [np.nan for _ in range(X.shape[0])]
impute = Imputer(strat_num="mean",
strat_cat="most_frequent",
min_frac_cols=0.5)
X, y = impute.fit_transform(X, y_bin)
assert len(X.columns) == 30 # Original number of columns
assert X.isna().sum().sum() == 0
def test_rows_too_many_nans():
"""Assert that rows with too many missing values are dropped."""
X = X_bin.copy()
for i in range(5): # Add 5 rows with all NaN values
X.loc[len(X)] = [np.nan for _ in range(X.shape[1])]
y = [np.random.randint(2) for _ in range(len(X))]
impute = Imputer(strat_num="mean",
strat_cat="most_frequent",
min_frac_rows=0.5)
X, y = impute.fit_transform(X, y)
assert len(X) == 569 # Original size
assert X.isna().sum().sum() == 0
def test_imputing_non_numeric_most_frequent():
"""Assert that the most_frequent strategy for non-numerical works."""
imputer = Imputer(strat_cat="most_frequent")
X, y = imputer.fit_transform(X10_sn, y10)
assert X.iloc[0, 2] == "d"
assert X.isna().sum().sum() == 0
def test_imputing_non_numeric_drop():
"""Assert that the drop strategy for non-numerical works."""
imputer = Imputer(strat_cat="drop")
X, y = imputer.fit_transform(X10_sn, y10)
assert len(X) == 9
assert X.isna().sum().sum() == 0
def test_imputing_non_numeric_string():
"""Assert that imputing a string for non-numerical values works."""
imputer = Imputer(strat_cat="missing")
X, y = imputer.fit_transform(X10_sn, y10)
assert X.iloc[0, 2] == "missing"
assert X.isna().sum().sum() == 0
def test_imputing_numeric_most_frequent():
"""Assert that imputing the most_frequent for numerical values works."""
imputer = Imputer(strat_num="most_frequent")
X, y = imputer.fit_transform(X10_nan, y10)
assert X.iloc[0, 0] == 3
assert X.isna().sum().sum() == 0
def test_imputing_numeric_mean():
"""Assert that imputing the mean for numerical values works."""
imputer = Imputer(strat_num="mean")
X, y = imputer.fit_transform(X10_nan, y10)
assert X.iloc[0, 0] == pytest.approx(2.577778, rel=1e-6, abs=1e-12)
assert X.isna().sum().sum() == 0
def test_imputing_numeric_number():
"""Assert that imputing a number for numerical values works."""
imputer = Imputer(strat_num=3.2)
X, y = imputer.fit_transform(X10_nan, y10)
assert X.iloc[0, 0] == 3.2
assert X.isna().sum().sum() == 0
def test_imputing_numeric_drop():
"""Assert that imputing drop for numerical values works."""
imputer = Imputer(strat_num="drop")
X, y = imputer.fit_transform(X10_nan, y10)
assert len(X) == 8
assert X.isna().sum().sum() == 0
def test_imputer_is_fitted():
"""Assert that an error is raised if the instance is not fitted."""
pytest.raises(NotFittedError, Imputer().transform, X_bin, y_bin)
def test_invalid_min_frac_cols():
"""Assert that an error is raised for invalid min_frac_cols."""
imputer = Imputer(min_frac_cols=5.2)
pytest.raises(ValueError, imputer.fit, X_bin, y_bin)
def test_strat_num_parameter():
"""Assert that the strat_num parameter is set correctly."""
imputer = Imputer(strat_num="invalid")
pytest.raises(ValueError, imputer.fit, X_bin, y_bin)
def test_load_data_with_no_trainer():
"""Assert that an error is raised when data is provided without a trainer."""
Imputer().save(FILE_DIR + "imputer")
pytest.raises(TypeError, ATOMLoader, FILE_DIR + "imputer", data=(X_bin, ))
