def test_as_np_vector_ndarray_2d(self):
p = SkProphet('date', True, [], {})
y = np.array([[1, 4], [2, 5], [3, 6]])
_y = p._as_np_vector(y)
self.assertIsInstance(_y, np.ndarray)
self.assertEqual(len(_y.shape), 1)
self.assertEqual(_y.tolist(), [1, 2, 3])
def test_as_np_vector_dataframe_2d(self):
p = SkProphet('date', True, [], {})
y = pd.DataFrame({'a': [1, 2, 3], 'b': [4, 5, 6]})
_y = p._as_np_vector(y)
self.assertIsInstance(_y, np.ndarray)
self.assertEqual(len(_y.shape), 1)
self.assertEqual(_y.tolist(), [1, 2, 3])
def test_as_np_vector_series(self):
p = SkProphet('date', True, [], {})
y = pd.Series([1, 2, 3])
_y = p._as_np_vector(y)
self.assertIsInstance(_y, np.ndarray)
self.assertEqual(len(_y.shape), 1)
self.assertEqual(_y.tolist(), [1, 2, 3])
def test_prophet_kwargs_set_params(self):
prophet_original_pars = {'n_changepoints': 10}
m = SkProphet(prophet_kwargs=prophet_original_pars)
prophet_new_pars = {
'prophet_kwargs': {
'yearly_seasonality': 4
},
'n_changepoints': 2,
}
m.set_params(**prophet_new_pars)
expected_prophet_kwargs = {
'yearly_seasonality': 4,
'n_changepoints': 2
}
self.assertEqual(m.prophet_kwargs, expected_prophet_kwargs)
self.assertEqual(m.yearly_seasonality, 4)
self.assertEqual(m.n_changepoints, 2)
def test_init_extra_regressors_dict(self):
sk_date_column = 'ds'
sk_yhat_only = True
sk_extra_regressors = [dict(name='x', mode='multiplicative')]
prophet_kwargs = {'daily_seasonality': True}
p = SkProphet(sk_date_column, sk_yhat_only, sk_extra_regressors,
prophet_kwargs)
# Extra regressors are set on initialization
self.assertTrue(p.extra_regressors)
def test_init_extra_regressors_type_error(self):
sk_date_column = 'ds'
sk_yhat_only = True
sk_extra_regressors = [4]
prophet_kwargs = {'daily_seasonality': True}
# Extra regressors of a bad type
with self.assertRaises(TypeError):
SkProphet(sk_date_column, sk_yhat_only, sk_extra_regressors,
prophet_kwargs)
def test_init_extra_regressors_str(self):
sk_date_column = 'ds'
sk_yhat_only = True
sk_extra_regressors = ['x']
prophet_kwargs = {'daily_seasonality': True}
p = SkProphet(sk_date_column, sk_yhat_only, sk_extra_regressors,
prophet_kwargs)
# Extra regressors are set on initialization
self.assertTrue(p.extra_regressors)
def test_init(self):
sk_date_column = 'ds'
sk_yhat_only = True
sk_extra_regressors = []
prophet_kwargs = {'daily_seasonality': True}
p = SkProphet(sk_date_column, sk_yhat_only, sk_extra_regressors,
prophet_kwargs)
# No extra regressors
self.assertTrue(p.daily_seasonality)
self.assertFalse(p.extra_regressors)
def test_repr(self):
# Initialize
sk_date_column = 'date'
sk_yhat_only = False
sk_extra_regressors = [dict(name='x', mode='multiplicative')]
prophet_kwargs = {'daily_seasonality': True}
skprophet = SkProphet(sk_date_column, sk_yhat_only,
sk_extra_regressors, prophet_kwargs)
# Assert: do not assert prophet internal attributes
expected = ('SkProphet('
'sk_date_column="date", '
'sk_yhat_only=False, '
'sk_extra_regressors=')
self.assertEqual(str(skprophet)[:len(expected)], expected)
def test_get_params_shallow(self):
sk_date_column = 'ds'
sk_yhat_only = True
sk_extra_regressors = [dict(name='x', mode='multiplicative')]
prophet_kwargs = {'daily_seasonality': True}
p = SkProphet(sk_date_column, sk_yhat_only, sk_extra_regressors,
prophet_kwargs)
# Assert shallow equality
params = p.get_params(deep=False)
self.assertEqual(len(params), 4)
self.assertEqual(params['sk_date_column'], sk_date_column)
self.assertEqual(params['sk_yhat_only'], sk_yhat_only)
self.assertEqual(params['sk_extra_regressors'], sk_extra_regressors)
self.assertEqual(
params['prophet_kwargs']['daily_seasonality'],
prophet_kwargs['daily_seasonality'],
)
# Assert shallow same objects
self.assertTrue(params['sk_date_column'] is sk_date_column)
self.assertTrue(params['sk_extra_regressors'] is sk_extra_regressors)
self.assertTrue(
params['sk_extra_regressors'][0] is sk_extra_regressors[0])
self.assertTrue(params['prophet_kwargs'] is prophet_kwargs)
def test_set_params(self):
sk_date_column = 'ds'
sk_yhat_only = True
sk_extra_regressors = [dict(name='x', mode='multiplicative')]
prophet_kwargs = {'daily_seasonality': True}
p = SkProphet(sk_date_column, sk_yhat_only, sk_extra_regressors,
prophet_kwargs)
# Alternative params
alternative_params = {
'sk_date_column': 'date',
'sk_extra_regressors': [dict(name='z')],
}
new_params = p.get_params(deep=True)
new_params.update(alternative_params)
# Assert params
p.set_params(**new_params)
params = p.get_params(deep=False)
self.assertEqual(new_params, params)
self.assertEqual(p.sk_date_column, 'date')
self.assertEqual(len(p.sk_extra_regressors), 1)
self.assertEqual(p.sk_extra_regressors[0]['name'], 'z')
def test_estimate_rename_ds(self):
p = SkProphet('date', True, [], {})
X = self._get_dataset()
X = X.rename({'ds': 'date'}, axis=1)
y_pred = p.fit(X).predict(X)
self.assertIsInstance(y_pred, np.ndarray)
def test_estimate_full_output(self):
p = SkProphet('ds', False, [], {})
X = self._get_dataset()
y_pred = p.fit(X).predict(X)
self.assertIsInstance(y_pred, pd.DataFrame)
def test_estimate(self):
p = SkProphet('ds', True, [], {})
X = self._get_dataset()
y_pred = p.fit(X).predict(X)
self.assertIsInstance(y_pred, np.ndarray)
def test_fit_X_not_dataframe_error(self):
p = SkProphet('ds', True, [], {})
X = [[1, 2], [3, 4]]
with self.assertRaises(TypeError):
p.fit(X)
def test_fit_y_list(self):
p = SkProphet('date', True, [], {})
X = self._get_dataset()
p2 = p.fit(X[['ds', 'x']], X.y.values.tolist())
self.assertEqual(p, p2)
def test_fit_y_str(self):
p = SkProphet('date', True, [], {})
X = self._get_dataset()
X = X.rename({'y': 'mica'}, axis=1)
p2 = p.fit(X, 'mica')
self.assertEqual(p, p2)
def test_fit_X_only_rename_ds(self):
p = SkProphet('date', True, [], {})
X = self._get_dataset()
X = X.rename({'ds': 'date'}, axis=1)
p2 = p.fit(X)
self.assertEqual(p, p2)
def test_fit_X_only(self):
p = SkProphet('ds', True, [], {})
X = self._get_dataset()
p2 = p.fit(X)
self.assertEqual(p, p2)
