def test_fit_with_horizon_and_more_rounds(self) -> type(None):
"""
Test `fit` method with horizon and several evaluational
rounds.
:return:
None
"""
candidates = {
MovingAverageForecaster(): [{
'rolling_kwargs': {
'window': w,
'min_periods': 1
}
} for w in range(1, 3)],
MovingMedianForecaster(): [{
'rolling_kwargs': {
'window': w,
'min_periods': 1
}
} for w in range(3, 4)]
}
selector = OnTheFlySelector(candidates,
horizon=3,
n_evaluational_rounds=3)
df = pd.DataFrame(
[[1, 2], [1, 3], [1, 6], [1, 5], [1, 4], [1, 1], [1, 3], [1, 2],
[2, 3], [2, 4], [2, 4.5], [2, 4], [2, 8], [2, 7], [2, 5], [2, 4]],
columns=['key', 'target'])
selector.fit(df, 'target', ['key'])
self.assertTrue(
isinstance(selector.best_scores_['forecaster'][1],
MovingMedianForecaster))
self.assertEquals(selector.best_scores_['forecaster'][1].get_params(),
{'rolling_kwargs': {
'window': 3,
'min_periods': 1
}})
self.assertEquals(selector.best_scores_['score'][1], -59 / 9)
self.assertTrue(
isinstance(selector.best_scores_['forecaster'][2],
MovingAverageForecaster))
self.assertEquals(selector.best_scores_['forecaster'][2].get_params(),
{'rolling_kwargs': {
'window': 2,
'min_periods': 1
}})
self.assertAlmostEqual(selector.best_scores_['score'][2],
-6.073784722222222)
def test_fit_with_scoring_keys(self) -> type(None):
"""
Test `fit` method with group-wise selection instead of
series-wise selection.
:return:
None
"""
candidates = {
MovingAverageForecaster(): [{
'rolling_kwargs': {
'window': w,
'min_periods': 1
}
} for w in range(1, 3)],
MovingMedianForecaster(): [{
'rolling_kwargs': {
'window': w,
'min_periods': 1
}
} for w in range(3, 4)]
}
selector = OnTheFlySelector(candidates)
df = pd.DataFrame(
[[1, 1, 2], [1, 1, 3], [1, 1, 6], [1, 1, 5], [1, 2, 3], [1, 2, 4],
[1, 2, 4.5], [1, 2, 1], [2, 1, 3], [2, 1, 2], [2, 1, -1],
[2, 1, 7], [2, 2, 0], [2, 2, 3], [2, 2, 2], [2, 2, 5]],
columns=['group', 'object', 'target'])
selector.fit(df, 'target', ['group', 'object'], scoring_keys=['group'])
self.assertTrue(
isinstance(selector.best_scores_['forecaster'][1],
MovingAverageForecaster))
self.assertEquals(selector.best_scores_['forecaster'][1].get_params(),
{'rolling_kwargs': {
'window': 2,
'min_periods': 1
}})
self.assertEquals(selector.best_scores_['score'][1], -5.40625)
self.assertTrue(
isinstance(selector.best_scores_['forecaster'][2],
MovingMedianForecaster))
self.assertEquals(selector.best_scores_['forecaster'][2].get_params(),
{'rolling_kwargs': {
'window': 3,
'min_periods': 1
}})
self.assertEquals(selector.best_scores_['score'][2], -17)
def test_predict(self) -> type(None):
"""
Test `predict` method.
:return:
None
"""
selector = OnTheFlySelector(horizon=2)
selector.best_scores_ = pd.DataFrame(
[[-1, MovingAverageForecaster({
'window': 2
}).fit(None)],
[-2, MovingMedianForecaster({
'window': 4
}).fit(None)]],
columns=['score', 'forecaster'],
index=[1, 2])
selector.best_scores_.index.name = 'key'
selector.name_of_target_ = 'target'
selector.scoring_keys_ = ['key']
selector.series_keys_ = ['key']
selector.evaluation_fn_ = None
df = pd.DataFrame(
[[1, 2], [1, 3], [1, 6], [1, 5], [2, 3], [2, 4], [2, 4.5], [2, 1]],
columns=['key', 'target'])
result = selector.predict(df)
true_answer = pd.DataFrame([[1, 5.5], [1, 5.25], [2, 3.5], [2, 3.75]],
columns=['key', 'prediction'],
index=[0, 1, 0, 1])
self.assertTrue(result.equals(true_answer))
def test_fit_with_evaluational_interval(self) -> type(None):
"""
Test `fit` method with selection based on results for
several steps.
:return:
None
"""
candidates = {
MovingAverageForecaster(): [{
'rolling_kwargs': {
'window': w,
'min_periods': 1
}
} for w in range(1, 3)],
MovingMedianForecaster(): [{
'rolling_kwargs': {
'window': w,
'min_periods': 1
}
} for w in range(3, 4)]
}
selector = OnTheFlySelector(candidates, n_evaluational_rounds=3)
df = pd.DataFrame([[1, 2], [1, 3], [1, 6], [1, 5], [1, 4], [1, 3],
[2, 3], [2, 4], [2, 4.5], [2, 1], [2, 7], [2, 2]],
columns=['key', 'target'])
selector.fit(df, 'target', ['key'])
self.assertTrue(
isinstance(selector.best_scores_['forecaster'][1],
MovingAverageForecaster))
self.assertEquals(selector.best_scores_['forecaster'][1].get_params(),
{'rolling_kwargs': {
'window': 1,
'min_periods': 1
}})
self.assertEqual(selector.best_scores_['score'][1], -1)
self.assertTrue(
isinstance(selector.best_scores_['forecaster'][2],
MovingMedianForecaster))
self.assertEquals(selector.best_scores_['forecaster'][2].get_params(),
{'rolling_kwargs': {
'window': 3,
'min_periods': 1
}})
self.assertAlmostEqual(selector.best_scores_['score'][2], -8.083333333)
def test_fit(self) -> type(None):
"""
Test `fit` method.
:return:
None
"""
candidates = {
MovingAverageForecaster(): [{
'rolling_kwargs': {
'window': w,
'min_periods': 1
}
} for w in range(1, 3)],
MovingMedianForecaster(): [{
'rolling_kwargs': {
'window': w,
'min_periods': 1
}
} for w in range(3, 4)]
}
selector = OnTheFlySelector(candidates)
df = pd.DataFrame(
[[1, 2], [1, 3], [1, 6], [1, 5], [2, 3], [2, 4], [2, 4.5], [2, 1]],
columns=['key', 'target'])
selector.fit(df, 'target', ['key'])
self.assertTrue(
isinstance(selector.best_scores_['forecaster'][1],
MovingAverageForecaster))
self.assertEquals(selector.best_scores_['forecaster'][1].get_params(),
{'rolling_kwargs': {
'window': 2,
'min_periods': 1
}})
self.assertEquals(selector.best_scores_['score'][1], -0.25)
self.assertTrue(
isinstance(selector.best_scores_['forecaster'][2],
MovingMedianForecaster))
self.assertEquals(selector.best_scores_['forecaster'][2].get_params(),
{'rolling_kwargs': {
'window': 3,
'min_periods': 1
}})
self.assertEquals(selector.best_scores_['score'][2], -9)
def test_that_fit_warns_on_series_of_unequal_length(self) -> type(None):
"""
Test that fit to a series of different length results in
`RuntimeWarning`.
:return:
None
"""
selector = OnTheFlySelector()
with warnings.catch_warnings(record=True) as caught_warnings:
df = pd.DataFrame(
[[1, 2], [1, 3], [1, 6], [2, 8], [2, 7], [2, 5], [2, 4]],
columns=['key', 'target'])
selector.fit(df, 'target', ['key'])
self.assertTrue(len(caught_warnings) == 1)
self.assertTrue(
issubclass(caught_warnings[-1].category, RuntimeWarning))
def test_fit_with_default_candidates(self) -> type(None):
"""
Test `fit` method with default set of candidates.
:return:
None
"""
selector = OnTheFlySelector()
df = pd.DataFrame(
[[1, 2], [1, 3], [1, 6], [1, 5], [1, 8], [1, 3], [1, 4], [1, 2],
[1, 5], [1, 7], [1, 5.5], [2, 3], [2, 4], [2, 4.5], [2, 1],
[2, 0], [2, 0], [2, 9], [2, 4], [2, 8], [2, 5], [2, 3]],
columns=['key', 'target'])
selector.fit(df, 'target', ['key'])
self.assertTrue(
isinstance(selector.best_scores_['forecaster'][1],
ExponentialMovingAverageForecaster))
self.assertEquals(
selector.best_scores_['forecaster'][1].get_params(), {
'ewm_kwargs': {
'halflife': 1.0,
'min_periods': 1
},
'n_steps_to_use': 10
})
self.assertAlmostEqual(selector.best_scores_['score'][1], -0.001978206)
self.assertTrue(
isinstance(selector.best_scores_['forecaster'][2],
MovingAverageForecaster))
self.assertEquals(selector.best_scores_['forecaster'][2].get_params(),
{'rolling_kwargs': {
'window': 10,
'min_periods': 1
}})
self.assertAlmostEqual(selector.best_scores_['score'][2], -0.7225)