def fallback(mase_all, smape_all, train, test, m):
naive = NaiveSeasonal(K=m)
naive.fit(train)
forecast = naive.predict(len(test))
mase_all.append(np.vstack([
mase_m4(train, test, forecast, m=m),
]))
smape_all.append(np.vstack([
smape_m4(test, forecast),
]))
def test_predict_ensemble_local_models(self):
naive = NaiveSeasonal(K=5)
theta = Theta()
naive_ensemble = NaiveEnsembleModel([naive, theta])
naive_ensemble.fit(self.series1 + self.series2)
forecast_naive_ensemble = naive_ensemble.predict(5)
naive.fit(self.series1 + self.series2)
theta.fit(self.series1 + self.series2)
forecast_mean = 0.5 * naive.predict(5) + 0.5 * theta.predict(5)
self.assertTrue(
np.array_equal(forecast_naive_ensemble.values(),
forecast_mean.values()))
def test_forecasting_residuals(self):
model = NaiveSeasonal(K=1)
# test zero residuals
constant_ts = ct(length=20)
residuals = model.residuals(constant_ts)
np.testing.assert_almost_equal(residuals.univariate_values(),
np.zeros(len(residuals)))
# test constant, positive residuals
linear_ts = lt(length=20)
residuals = model.residuals(linear_ts)
np.testing.assert_almost_equal(np.diff(residuals.univariate_values()),
np.zeros(len(residuals) - 1))
np.testing.assert_array_less(np.zeros(len(residuals)),
residuals.univariate_values())
def naive2_groe(ts: TimeSeries, n: int, m: int):
"""
Return the prediction of the naive2 baseline
"""
# It will be better to use R functions
ts_des = ts
seasonOut = 1
if m > 1:
if check_seasonality(ts, m=int(m), max_lag=2 * m):
_, season = extract_trend_and_seasonality(ts, m, model=ModelMode.MULTIPLICATIVE)
ts_des = remove_from_series(ts, season, model=ModelMode.MULTIPLICATIVE)
seasonOut = season[-m:].shift(m)
seasonOut = seasonOut.append_values(seasonOut.values())[:n]
naive2 = NaiveSeasonal(K=1)
naive2.fit(ts_des)
return naive2.predict(n) * seasonOut
def test_fit_univar_ts_with_covariates_for_local_models(self):
naive = NaiveEnsembleModel(
[NaiveDrift(),
NaiveSeasonal(),
Theta(),
ExponentialSmoothing()])
with self.assertRaises(ValueError):
naive.fit(self.series1, self.series2)
def test_input_models_local_models(self):
with self.assertRaises(ValueError):
NaiveEnsembleModel([])
with self.assertRaises(ValueError):
NaiveEnsembleModel(
[NaiveDrift, NaiveSeasonal, Theta, ExponentialSmoothing])
with self.assertRaises(ValueError):
NaiveEnsembleModel(
[NaiveDrift(), NaiveSeasonal,
Theta(),
ExponentialSmoothing()])
NaiveEnsembleModel(
[NaiveDrift(),
NaiveSeasonal(),
Theta(),
ExponentialSmoothing()])
m = int(info_dataset.Frequency[cat[0] + "1"])
for train, test in _build_tqdm_iterator(zip(ts_train, ts_test),
verbose=True):
train_des = train
seasonOut = 1
if m > 1:
if check_seasonality(train, m=m, max_lag=2 * m):
_, season = extract_trend_and_seasonality(
train, m, model=ModelMode.MULTIPLICATIVE)
train_des = remove_from_series(
train, season, model=ModelMode.MULTIPLICATIVE)
seasonOut = season[-m:].shift(m)
seasonOut = seasonOut.append_values(seasonOut.values())
seasonOut = seasonOut[:len(test)]
naive = NaiveDrift()
naive2 = NaiveSeasonal(K=1)
naiveSeason = NaiveSeasonal(K=m)
ses = ExponentialSmoothing(trend=None,
seasonal=None,
seasonal_periods=m)
holt = ExponentialSmoothing(seasonal=None,
damped=False,
trend='additive',
seasonal_periods=m)
damp = ExponentialSmoothing(seasonal=None,
damped=True,
trend='additive',
seasonal_periods=m)
naive.fit(train)
naive2.fit(train_des)
naiveSeason.fit(train)
(ExponentialSmoothing(), 5.6),
(ARIMA(12, 2, 1), 10),
(ARIMA(1, 1, 1), 40),
(StatsForecastAutoARIMA(period=12), 4.8),
(Croston(version="classic"), 34),
(Croston(version="tsb", alpha_d=0.1, alpha_p=0.1), 34),
(Theta(), 11.3),
(Theta(1), 20.2),
(Theta(-1), 9.8),
(FourTheta(1), 20.2),
(FourTheta(-1), 9.8),
(FourTheta(trend_mode=TrendMode.EXPONENTIAL), 5.5),
(FourTheta(model_mode=ModelMode.MULTIPLICATIVE), 11.4),
(FourTheta(season_mode=SeasonalityMode.ADDITIVE), 14.2),
(FFT(trend="poly"), 11.4),
(NaiveSeasonal(), 32.4),
(KalmanForecaster(dim_x=3), 17.0),
]
if TORCH_AVAILABLE:
models += [
(LinearRegressionModel(lags=12), 11.0),
(RandomForest(lags=12, n_estimators=200, max_depth=3), 15.5),
]
# forecasting models with exogenous variables support
multivariate_models = [
(VARIMA(1, 0, 0), 55.6),
(VARIMA(1, 1, 1), 57.0),
(KalmanForecaster(dim_x=30), 30.0),
]
smape_all = []
m = int(info_dataset.Frequency[freq[0] + '1'])
for train, test in _build_tqdm_iterator(zip(ts_train, ts_test), verbose=True):
# remove seasonality
train_des = train
seasonOut = 1
season = constant_timeseries(length=len(train), freq=train.freq_str, start_ts=train.start_time())
if m > 1:
if check_seasonality(train, m=m, max_lag=2 * m):
pass
_, season = extract_trend_and_seasonality(train, m, model=ModelMode.MULTIPLICATIVE)
train_des = remove_from_series(train, season, model=ModelMode.MULTIPLICATIVE)
seasonOut = season[-m:].shift(m)
seasonOut = seasonOut.append_values(seasonOut.values())[:len(test)]
# model selection
naiveSeason = NaiveSeasonal(K=m)
naive2 = NaiveSeasonal(K=1)
ses = ExponentialSmoothing(trend=None, seasonal=None, seasonal_periods=m)
holt = ExponentialSmoothing(seasonal=None, damped=False, trend='additive', seasonal_periods=m)
damp = ExponentialSmoothing(seasonal=None, damped=True, trend='additive', seasonal_periods=m)
fourtheta = FourTheta.select_best_model(train, [1, 2, 3], m)
theta = Theta(theta=0, season_mode=SeasonalityMode.MULTIPLICATIVE, seasonality_period=m)
models_simple = [naiveSeason, theta, fourtheta]
models_des = [naive2, ses, holt, damp]
# Linear Regression (with constraints)
def train_pred(id_start=None, id_end=None):
for m in models_simple:
def get_local_models(self):
return [NaiveDrift(), NaiveSeasonal(5), NaiveSeasonal(10)]
def test_call_predict_local_models(self):
naive_ensemble = NaiveEnsembleModel([NaiveSeasonal(), Theta()])
with self.assertRaises(Exception):
naive_ensemble.predict(5)
naive_ensemble.fit(self.series1)
naive_ensemble.predict(5)
seasonOut = 1
season = constant_timeseries(length=len(train),
freq=train.freq_str(),
start_ts=train.start_time())
if m > 1:
if check_seasonality(train, m=m, max_lag=2 * m):
pass
_, season = extract_trend_and_seasonality(
train, m, model=ModelMode.MULTIPLICATIVE)
train_des = remove_from_series(
train, season, model=ModelMode.MULTIPLICATIVE)
seasonOut = season[-m:].shift(m)
seasonOut = seasonOut.append_values(
seasonOut.values())[:len(test)]
# model selection
naiveSeason = NaiveSeasonal(K=m)
naive2 = NaiveSeasonal(K=1)
ses = ExponentialSmoothing(trend=None,
seasonal=None,
seasonal_periods=m)
holt = ExponentialSmoothing(seasonal=None,
damped=False,
trend='additive',
seasonal_periods=m)
damp = ExponentialSmoothing(seasonal=None,
damped=True,
trend='additive',
seasonal_periods=m)
fourtheta = FourTheta.select_best_model(train, [1, 2, 3], m)
theta = Theta(theta=0,