def test_df_cat_and_num_variables_roc_auc(df_test_num_cat):
X, y = df_test_num_cat
sel = SelectByTargetMeanPerformance(
variables=None,
scoring="roc_auc_score",
threshold=0.6,
bins=3,
strategy="equal_width",
cv=2,
random_state=1,
)
sel.fit(X, y)
# expected result
Xtransformed = X[["var_A", "var_B"]]
performance_dict = {"var_A": 0.841, "var_B": 0.776, "var_C": 0.481, "var_D": 0.496}
# test init params
assert sel.variables == list(X.columns)
assert sel.scoring == "roc_auc_score"
assert sel.threshold == 0.60
assert sel.cv == 2
assert sel.random_state == 1
# test fit attrs
assert sel.variables_categorical_ == ["var_A", "var_B"]
assert sel.variables_numerical_ == ["var_C", "var_D"]
assert sel.selected_features_ == ["var_A", "var_B"]
assert all(
np.round(sel.feature_performance_[f], 3) == performance_dict[f]
for f in sel.feature_performance_.keys()
)
# test transform output
pd.testing.assert_frame_equal(sel.transform(X), Xtransformed)
def test_categorical_variables_roc_auc(df_test_num_cat):
X, y = df_test_num_cat
X = X[["var_A", "var_B"]]
sel = SelectByTargetMeanPerformance(
variables=None,
scoring="roc_auc_score",
threshold=0.78,
cv=2,
random_state=1,
)
sel.fit(X, y)
# expected result
Xtransformed = X["var_A"].to_frame()
# performance_dict = {"var_A": 0.841, "var_B": 0.776}
# test init params
assert sel.variables is None
assert sel.scoring == "roc_auc_score"
assert sel.threshold == 0.78
assert sel.cv == 2
assert sel.random_state == 1
# test fit attrs
assert sel.variables_ == list(X.columns)
assert sel.variables_categorical_ == list(X.columns)
assert sel.variables_numerical_ == []
assert sel.features_to_drop_ == ["var_B"]
# assert all(
# np.round(sel.feature_performance_[f], 3) == performance_dict[f]
# for f in sel.feature_performance_.keys()
# )
# test transform output
pd.testing.assert_frame_equal(sel.transform(X), Xtransformed)