def test_na_vals_reg_plot_raise_warning():
X, y = load_diabetes(return_X_y=True)
X = pd.DataFrame(X)
y[::50] = np.NaN
X['target_col'] = y
scatter_alpha = _get_scatter_alpha('auto', X['target_col'])
scatter_size = _get_scatter_size('auto', X['target_col'])
with pytest.warns(UserWarning,
match="Missing values in target_col have "
"been removed for regression"):
plot(X, 'target_col')
with pytest.warns(UserWarning,
match="Missing values in target_col have "
"been removed for regression"):
plot_regression_continuous(X,
'target_col',
scatter_alpha=scatter_alpha,
scatter_size=scatter_size)
with pytest.warns(UserWarning,
match="Missing values in target_col have "
"been removed for regression"):
plot_regression_categorical(X,
'target_col',
scatter_alpha=scatter_alpha,
scatter_size=scatter_size)
def test_plot_regression_continuous_with_target_outliers():
df = pd.DataFrame(
data={
"feature": np.random.randint(low=1, high=100, size=200),
# target values are bound between 50 and 100
"target": np.random.randint(low=50, high=100, size=200)
})
# append single outlier record with target value 0
df = df.append({"feature": 50, "target": 0}, ignore_index=True)
with pytest.warns(UserWarning,
match="Dropped 1 outliers in column target."):
plot_regression_continuous(df, 'target')
def test_plot_wrong_target_type():
X, y = make_blobs()
X = pd.DataFrame(X)
X['target'] = y
with pytest.raises(ValueError, match="need continuous"):
plot_regression_categorical(X, 'target')
with pytest.raises(ValueError, match="need continuous"):
plot_regression_continuous(X, 'target')
X['target'] = X[0]
with pytest.raises(ValueError, match="need categorical"):
plot_classification_categorical(X, 'target')
with pytest.raises(ValueError, match="need categorical"):
plot_classification_continuous(X, 'target')
def test_label_truncation():
a = ('a_really_long_name_that_would_mess_up_the_layout_a_lot'
'_by_just_being_very_long')
b = ('the_target_that_has_an_equally_long_name_which_would_'
'mess_up_everything_as_well_but_in_different_places')
df = pd.DataFrame({a: np.random.uniform(0, 1, 1000)})
df[b] = df[a] + np.random.uniform(0, 0.1, 1000)
res = plot_regression_continuous(df, target_col=b)
assert res[0, 0].get_ylabel() == 'the_target_that_h...'
assert res[0, 0].get_xlabel() == 'a_really_long_nam...'
set_config(truncate_labels=False)
res = plot_regression_continuous(df, target_col=b)
assert res[0, 0].get_ylabel() == b
assert res[0, 0].get_xlabel() == a
set_config(truncate_labels=True)
def test_plot_regression_with_target_outliers():
df = pd.DataFrame(
data={
"feature": np.random.randint(low=1, high=100, size=200),
# target values are bound between 50 and 100
"target": np.random.randint(low=50, high=100, size=200)
})
# append single outlier record with target value 0
df = df.append({"feature": 50, "target": 0}, ignore_index=True)
with pytest.warns(UserWarning,
match="Dropped 1 outliers in column target."):
plot_regression_continuous(df, target_col='target')
with pytest.warns(UserWarning,
match="Dropped 1 outliers in column target."):
plot_regression_categorical(df, target_col='target')
res = plot(df, target_col='target')
assert len(res) == 3
ax = res[0]
# ensure outlier at 0 was removed
assert ax.get_xticks()[0] == 40
