def test_supervised_data_split_sizes():
# Test that the data splits are of correct sizes
toy_data = pd.DataFrame({
"col1": [1, 1, 1, 1],
"col2": [2, 2, 2, 2],
"col3": [3, 3, 3, 3]
})
supervised_data = pymleda.SupervisedData(toy_data,
x_cols=["col1", "col2"],
y_cols=["col3"])
assert len(supervised_data.test_df) == 0.25 * len(toy_data)
assert len(supervised_data.train_df) == 0.75 * len(toy_data)
# Reconstruct the original dataframe from the two subsets as a second check
reconstructed = pd.concat(
[supervised_data.test_df, supervised_data.train_df])
assert (pd.testing.assert_frame_equal(
reconstructed,
toy_data,
check_like=True,
) is None)
def test_supervised_data_invalid_input():
# Test that an Exception is raised with invalid input
toy_data = pd.DataFrame({
"col1": [1, 1, 1, 1],
"col2": [2, 2, 2, 2],
"col3": [3, 3, 3, 3]
})
with pytest.raises(Exception):
pymleda.SupervisedData(1, x_cols=["col1", "col2"], y_cols=["col3"])
with pytest.raises(Exception):
pymleda.SupervisedData(toy_data, x_cols=1, y_cols=["col3"])
with pytest.raises(Exception):
pymleda.SupervisedData(toy_data, x_cols=["col1", "col2"], y_cols=1)
def test_supervised_data_original_dataset():
# Simple test to ensure that the data attribute returns the original
# data unchanged
toy_data = pd.DataFrame({
"col1": [1, 1, 1, 1],
"col2": [2, 2, 2, 2],
"col3": [3, 3, 3, 3]
})
assert (pd.testing.assert_frame_equal(
pymleda.SupervisedData(
toy_data, x_cols=["col1", "col2"], y_cols=["col3"]).data,
toy_data,
) is None)
def test_supervised_data_y():
# Test that the y portions of the data only contain the y columns
toy_data = pd.DataFrame({
"col1": [1, 1, 1, 1],
"col2": [2, 2, 2, 2],
"col3": [3, 3, 3, 3]
})
supervised_data = pymleda.SupervisedData(toy_data,
x_cols=["col1", "col2"],
y_cols=["col3"])
assert list(supervised_data.y_train.columns) == ["col3"]
assert list(supervised_data.y_test.columns) == ["col3"]
def test_autoimpute_na_3(model_df):
"""Test that the input of the pymleda.autoimpute_na()
function is not a dataframe"""
with pytest.raises(Exception):
pymleda.SupervisedData(pymleda.autoimpute_na(777), model_df)