def test_dataframe_model(self, mimic_explainer, sample_cnt_per_grain, grains_dict):
X, _ = create_timeseries_data(sample_cnt_per_grain, 'time', 'y', grains_dict)
model = DataFrameTestModel(X.copy())
model = Pipeline([('test', model)])
features = list(X.columns.values) + list(X.index.names)
model_task = ModelTask.Unknown
kwargs = {'explainable_model_args': {'n_jobs': 1}, 'augment_data': False, 'reset_index': True}
if grains_dict:
kwargs['categorical_features'] = ['fruit']
mimic_explainer(model, X, LGBMExplainableModel, features=features, model_task=model_task, **kwargs)
def test_timestamp_featurization(self, sample_cnt_per_grain, grains_dict):
# create timeseries data
X, _ = create_timeseries_data(sample_cnt_per_grain, 'time', 'y',
grains_dict)
original_cols = list(X.columns.values)
# featurize and validate the timestamp column
featurizer = CustomTimestampFeaturizer(original_cols).fit(X)
result = featurizer.transform(X)
# Form a temporary dataframe for validation
tmp_result = pd.DataFrame(result)
# Assert there are no timestamp columns
assert ([
column for column in tmp_result.columns
if is_datetime(tmp_result[column])
] == [])
# Assert we have the expected number of columns - 1 time columns * 6 featurized plus original
assert (result.shape[1] == len(original_cols) + 6)