def test_model_comparison_give_non_null_performance_with_regression_and_categorical_feature(
):
# Given
cross_validation_n_folds = 2
features = pd.DataFrame({
"string_feature":
list(
np.random.choice(["Paris", "London", "Madrid", "Roma"],
n_samples - 1)) + [None],
})
comparison_dataset = ComparisonDataset(TaskName.REGRESSION, features,
numerical_target,
cross_validation_n_folds)
# When
model_comparison = TunedModelComparison(comparison_dataset,
max_parameters_to_test_in_tuning=5,
early_stopping_patience=1)
comparison = model_comparison.get_models_scores_and_training_time()
# Then
for model_name, performance_and_training_time in comparison.items():
performance = performance_and_training_time[MODEL_SCORE]
assert_that(~np.isnan(performance),
reason=f"Null performance value for model {model_name}")
def test_model_comparison_give_non_null_performance_with_classification():
# Given
features = pd.DataFrame(
{"numeric_feature": np.random.normal(size=n_samples)})
comparison_dataset = ComparisonDataset(TaskName.CLASSIFICATION, features,
categorical_target,
cross_validation_n_folds)
# When
comparison = ModelComparison(
comparison_dataset).get_models_scores_and_training_time()
# Then
for model_name, performance_and_training_time in comparison.items():
performance = performance_and_training_time[MODEL_SCORE]
assert_that(~np.isnan(performance),
reason=f"Null performance value for model {model_name}")
def test_model_comparison_give_non_null_performance_with_regression_and_numerial_feature(
):
# Given
cross_validation_n_folds = 2
features = pd.DataFrame(
{"numeric_feature": np.random.normal(size=n_samples)})
comparison_dataset = ComparisonDataset(TaskName.REGRESSION, features,
numerical_target,
cross_validation_n_folds)
# When
model_comparison = TunedModelComparison(comparison_dataset,
max_parameters_to_test_in_tuning=5,
early_stopping_patience=1)
comparison = model_comparison.get_models_scores_and_training_time()
# Then
for model_name, performance_and_training_time in comparison.items():
performance = performance_and_training_time[MODEL_SCORE]
assert_that(~np.isnan(performance),
reason=f"Null performance value for model {model_name}")
def test_model_comparison_give_non_null_performance_and_categorical_feature():
# Given
features = pd.DataFrame({
"string_feature":
np.random.choice(["Paris", "London", "Madrid", "Roma"], n_samples),
"numeric_feature":
np.random.normal(size=n_samples)
})
comparison_dataset = ComparisonDataset(TaskName.REGRESSION, features,
categorical_target,
cross_validation_n_folds)
# When
comparison = ModelComparison(
comparison_dataset).get_models_scores_and_training_time()
# Then
for model_name, performance_and_training_time in comparison.items():
performance = performance_and_training_time[MODEL_SCORE]
assert_that(~np.isnan(performance),
reason=f"Null performance value for model {model_name}")