def test_fit_target_to_feature_dict_working(self, data, X_y, estimator):
X, y = X_y
X_train, y_train, X_test, y_test = FeatureSplitter().transform(X, y)
target_to_feature_dict = data.draw(
str_target_to_feature_dicts(targets=y.columns, features=X.columns)
)
multi_feature_gar = MultiFeatureGAR(estimator)
multi_feature_gar.fit(X_train, y_train, target_to_features_dict=target_to_feature_dict)
def test_predict_has_explainers(self, forecaster, explainer, X_y):
X, y = X_y
X_train, y_train, X_test, y_test = FeatureSplitter().transform(X, y)
model = forecaster(LinearRegression(), explainer_type=explainer)
model.fit(X_train, y_train)
model.predict(X_test.iloc[:1, :])
assert len(model.estimators_) == y_test.shape[1]
for estimator in model.estimators_:
assert len(estimator.explainer_.explanations_) == 1
def test_explanations_columns(self, forecaster, explainer, X_y):
X, y = X_y
X_train, y_train, X_test, y_test = FeatureSplitter().transform(X, y)
model = forecaster(LinearRegression(), explainer_type=explainer)
model.fit(X_train, y_train)
model.predict(X_test.iloc[:1, :])
for target_column, explanation in model.explanations_.items():
assert set(explanation.columns).issuperset(set(X_train.columns))
assert set(model.explanations_.keys()) == set(y_train.columns)
def test_predict_target_to_feature_dict_explainable(
self, data, X_y, estimator):
X, y = X_y
X_train, y_train, X_test, y_test = FeatureSplitter().transform(X, y)
target_to_feature_dict = data.draw(
str_target_to_feature_dicts(targets=y.columns, features=X.columns))
multi_feature_gar = MultiFeatureGAR(estimator, explainer_type="shap")
multi_feature_gar.target_to_features_dict = target_to_feature_dict
multi_feature_gar.fit(X_train, y_train)
multi_feature_gar.predict(X_test)
def test_fit_target_to_feature_dict_consistent(self, data, X_y, estimator):
X, y = X_y
X_train, y_train, X_test, y_test = FeatureSplitter().transform(X, y)
target_to_feature_dict = data.draw(
str_target_to_feature_dicts(targets=y.columns, features=X.columns)
)
multi_feature_gar = MultiFeatureGAR(estimator)
multi_feature_gar.fit(X_train, y_train, target_to_features_dict=target_to_feature_dict)
for i, estimator_ in enumerate(multi_feature_gar.estimators_):
expected_n_features = len(target_to_feature_dict[y.columns[i]])
assert len(estimator_.coef_) == expected_n_features
def test_error_predict_target_to_feature_dict_wrong_X_shape(
self, data, X_y, estimator):
X, y = X_y
X_train, y_train, X_test, y_test = FeatureSplitter().transform(X, y)
target_to_feature_dict = data.draw(
str_target_to_feature_dicts(targets=y.columns, features=X.columns))
multi_feature_gar = MultiFeatureGAR(
estimator, target_to_features_dict=target_to_feature_dict)
multi_feature_gar.fit(X_train, y_train)
X_test = X_test.iloc[:, :2]
with pytest.raises(ValueError):
multi_feature_gar.predict(X_test)
def forecast_input(draw, max_lenth):
length = draw(st.integers(min_value=2, max_value=max_lenth))
horizon = draw(st.integers(min_value=1, max_value=length - 1))
X = draw(
giotto_time_series(
min_length=length,
max_length=max_lenth,
allow_nan=False,
allow_infinity=False,
))
y = horizon_shift(X, horizon=horizon)
X_train, y_train, X_test, y_test = FeatureSplitter().transform(X, y)
return X_train, y_train, X_test
def test_correct_fit_date(self, X_y):
base_model = LinearRegression()
feature_splitter = FeatureSplitter()
x, y = X_y[0], X_y[1]
x_train, y_train, x_test, y_test = feature_splitter.transform(x, y)
gar_no_feedforward = GAR(estimator=base_model)
gar_no_feedforward.fit(x_train, y_train)
predictions = gar_no_feedforward.predict(x_test)
assert len(predictions) == len(x_test)
np.testing.assert_array_equal(predictions.index, x_test.index)
gar_with_feedforward = GARFF(estimator=base_model)
gar_with_feedforward.fit(x_train, y_train)
predictions = gar_with_feedforward.predict(x_test)
assert len(predictions) == len(x_test)
np.testing.assert_array_equal(predictions.index, x_test.index)
def test_fit_as_multi_output_regressor_if_target_to_feature_none(
self, estimator, X_y
):
X, y = X_y
X_train, y_train, X_test, y_test = FeatureSplitter().transform(X, y)
multi_feature_gar = MultiFeatureGAR(estimator)
multi_feature_gar.fit(X_train, y_train)
gar = GAR(estimator)
gar.fit(X_train, y_train)
pd.testing.assert_frame_equal(
multi_feature_gar.predict(X_test), gar.predict(X_test),
)
def _split_train_test(self, X: pd.DataFrame, y: pd.DataFrame):
feature_splitter = FeatureSplitter()
return feature_splitter.transform(X, y)
def _split_train_test(self, X, y):
feature_splitter = FeatureSplitter()
return feature_splitter.transform(X, y)
