def test_ts_forecasting_lagged_data_operation():
train_input, predict_input, y_test = get_time_series()
node_lagged = PrimaryNode('lagged')
node_ridge = SecondaryNode('ridge', nodes_from=[node_lagged])
pipeline = Pipeline(node_ridge)
pipeline.fit_from_scratch(train_input)
predicted_output = pipeline.predict(predict_input)
predicted = np.ravel(predicted_output.predict)
assert len(predicted) == len(np.ravel(y_test))
def test_ts_forecasting_smoothing_data_operation():
train_input, predict_input, y_test = get_time_series()
model_names, _ = OperationTypesRepository().operations_with_tag(
tags=['smoothing'])
for smoothing_operation in model_names:
node_smoothing = PrimaryNode(smoothing_operation)
node_lagged = SecondaryNode('lagged', nodes_from=[node_smoothing])
node_ridge = SecondaryNode('ridge', nodes_from=[node_lagged])
pipeline = Pipeline(node_ridge)
pipeline.fit_from_scratch(train_input)
predicted_output = pipeline.predict(predict_input)
predicted = np.ravel(predicted_output.predict)
assert len(predicted) == len(np.ravel(y_test))
def test_classification_data_operations():
train_input, predict_input, y_test = get_small_classification_dataset()
model_names, _ = OperationTypesRepository().suitable_operation(
task_type=TaskTypesEnum.classification)
for data_operation in model_names:
node_data_operation = PrimaryNode(data_operation)
node_final = SecondaryNode('logit', nodes_from=[node_data_operation])
pipeline = Pipeline(node_final)
# Fit and predict for pipeline
pipeline.fit_from_scratch(train_input)
predicted_output = pipeline.predict(predict_input)
predicted = predicted_output.predict
assert len(predicted) == len(y_test)
def test_inf_and_nan_absence_after_pipeline_fitting_from_scratch():
train_input = get_nan_inf_data()
model_names, _ = OperationTypesRepository().suitable_operation(
task_type=TaskTypesEnum.regression)
for data_operation in model_names:
node_data_operation = PrimaryNode(data_operation)
node_final = SecondaryNode('linear', nodes_from=[node_data_operation])
pipeline = Pipeline(node_final)
# Fit and predict for pipeline
pipeline.fit_from_scratch(train_input)
predicted_output = pipeline.predict(train_input)
predicted = predicted_output.predict
assert np.sum(np.isinf(predicted)) == 0
assert np.sum(np.isnan(predicted)) == 0
def test_data_merge_in_pipeline():
""" Test check is the pipeline can correctly work with dynamic changes in
tables during the fit process
"""
# ridge
# / \ (merge operation)
# | ransac_lin_reg (remove several lines in table)
# \ /
# scaling
node_scaling = PrimaryNode('scaling')
node_lin_ransac = SecondaryNode('ransac_lin_reg', nodes_from=[node_scaling])
node_final = SecondaryNode('ridge', nodes_from=[node_lin_ransac, node_scaling])
pipeline = Pipeline(node_final)
features_options = {'informative': 2, 'bias': 2.0}
x_train, y_train, x_test, y_test = get_regression_dataset(features_options=features_options,
samples_amount=100,
features_amount=5)
# Define regression task
task = Task(TaskTypesEnum.regression)
# Prepare data to train the model
train_input = InputData(idx=np.arange(0, len(x_train)),
features=x_train,
target=y_train,
task=task,
data_type=DataTypesEnum.table)
# Fit and predict
pipeline.fit_from_scratch(train_input)
prediction = pipeline.predict(train_input)
assert prediction is not None
