def test_create_scoring():
classifier = RandomForestClassifier()
regressor = RandomForestRegressor()
res = create_scoring(classifier, "accuracy")
assert isinstance(res, OrderedDict)
assert "accuracy" in res
for k, v in res.items():
assert callable(v)
res = create_scoring(regressor, "neg_mean_squared_error")
assert isinstance(res, OrderedDict)
assert "neg_mean_squared_error" in res
for k, v in res.items():
assert callable(v)
res = create_scoring(
regressor, ["neg_mean_squared_error", "neg_median_absolute_error"])
assert isinstance(res, OrderedDict)
assert "neg_mean_squared_error" in res
assert "neg_median_absolute_error" in res
for k, v in res.items():
assert callable(v)
res = create_scoring(regressor, res)
assert isinstance(res, OrderedDict)
assert "neg_mean_squared_error" in res
assert "neg_median_absolute_error" in res
for k, v in res.items():
assert callable(v)
assert type(v) == type(res[k])
res = create_scoring(regressor, {"scorer1": "accuracy"})
assert isinstance(res, OrderedDict)
assert "scorer1" in res
for k, v in res.items():
assert callable(v)
assert type(v) == type(res[k])
res = create_scoring(regressor, {"scorer1": SCORERS["accuracy"]})
assert isinstance(res, OrderedDict)
assert "scorer1" in res
for k, v in res.items():
assert callable(v)
assert type(v) == type(res[k])
res = create_scoring(regressor, SCORERS["accuracy"])
assert isinstance(res, OrderedDict)
for k, v in res.items():
assert callable(v)
assert type(v) == type(res[k])
res = create_scoring(regressor, None)
assert "default_score" in res
assert isinstance(res, OrderedDict)
for k, v in res.items():
assert callable(v)
assert type(v) == type(res[k])
def test_approx_cross_validation_BoxCoxTargetTransformer():
np.random.seed(123)
X = np.random.randn(100, 10)
y = np.exp(np.random.randn(100))
for ll in (0, 0.1, 0.5, 2):
# Scorer entered as a string #
bb = BoxCoxTargetTransformer(Ridge(), ll=ll)
cv_res1, yhat1 = bb.approx_cross_validation(
X,
y,
scoring=["neg_mean_squared_error"],
cv=10,
return_predict=True)
assert isinstance(cv_res1, pd.DataFrame)
assert cv_res1.shape[0] == 10
assert "test_neg_mean_squared_error" in cv_res1
assert "train_neg_mean_squared_error" in cv_res1
assert yhat1.ndim == 1
assert yhat1.shape[0] == y.shape[0]
with pytest.raises(NotFittedError):
bb.predict(X)
with pytest.raises(NotFittedError):
bb.model.predict(X)
#########################################
### Scorer entered as a dictionnary ###
#########################################
scoring = create_scoring(Ridge(), ["neg_mean_squared_error"])
cv_res2, yhat2 = bb.approx_cross_validation(X,
y,
scoring=scoring,
cv=10,
return_predict=True)
assert isinstance(cv_res2, pd.DataFrame)
assert cv_res2.shape[0] == 10
assert "test_neg_mean_squared_error" in cv_res2
assert "train_neg_mean_squared_error" in cv_res2
assert yhat2.ndim == 1
assert yhat2.shape[0] == y.shape[0]
with pytest.raises(NotFittedError):
bb.predict(X)
with pytest.raises(NotFittedError):
bb.model.predict(X)
assert np.abs(cv_res2["test_neg_mean_squared_error"] -
cv_res1["test_neg_mean_squared_error"]).max() <= 10**(-5)
assert np.abs(cv_res2["train_neg_mean_squared_error"] -
cv_res1["train_neg_mean_squared_error"]).max() <= 10**(
-5)
assert np.max(np.abs(yhat2 - yhat1)) <= 10**(-5)