def test_make_future_dataframe_length(self):
mp = multi_prophet.MultiProphet(columns=PREDICTOR_COLUMNS)
mp.fit(self.df)
future_df = mp.make_future_dataframe(7)
initial_len = len(self.df)
self.assertEqual(initial_len + 7, len(future_df))
def test_add_country_holidays_all_models(self):
mp = multi_prophet.MultiProphet(columns=PREDICTOR_COLUMNS)
mp.add_country_holidays("US")
mp.fit(self.df)
future_df = mp.make_future_dataframe(7)
for model in mp.model_pool.values():
self.assertEqual(14, len(model.prophet.train_holiday_names))
def test_predict(self):
mp = multi_prophet.MultiProphet(columns=PREDICTOR_COLUMNS)
mp.fit(self.df)
future_df = mp.make_future_dataframe(7)
forecast = mp.predict(future_df)
self.assertEqual(2, len(forecast))
self.assertIsInstance(forecast["y"], pd.DataFrame)
self.assertIsInstance(forecast["y1"], pd.DataFrame)
def test_components_plot(self):
mp = multi_prophet.MultiProphet(columns=PREDICTOR_COLUMNS)
mp.fit(self.df)
future_df = mp.make_future_dataframe(7)
forecasts = mp.predict(future_df)
plots = mp.plot_components(forecasts).values()
for plot in plots:
self.assertIsInstance(plot, matplotlib.figure.Figure)
def test_add_regressor_all_models(self):
mp = multi_prophet.MultiProphet(columns=PREDICTOR_COLUMNS)
mp.add_regressor("Matchday")
for model in mp.model_pool.values():
extra_regressor = model.prophet.extra_regressors["Matchday"]
self.assertEqual(10.0, extra_regressor["prior_scale"])
self.assertEqual("auto", extra_regressor["standardize"])
self.assertEqual(0.0, extra_regressor["mu"])
self.assertEqual(1.0, extra_regressor["std"])
self.assertEqual("additive", extra_regressor["mode"])
def test_add_seasonality_all_models(self):
mp = multi_prophet.MultiProphet(columns=PREDICTOR_COLUMNS)
mp.add_seasonality(name="monthly", period=30.5, fourier_order=5)
for model in mp.model_pool.values():
seasonality = model.prophet.seasonalities["monthly"]
self.assertEqual(30.5, seasonality["period"])
self.assertEqual(5, seasonality["fourier_order"])
self.assertEqual(10, seasonality["prior_scale"])
self.assertEqual("additive", seasonality["mode"])
self.assertIsNone(seasonality["condition_name"])
def test_add_country_holidays_single_model(self):
mp = multi_prophet.MultiProphet(columns=PREDICTOR_COLUMNS)
mp.add_country_holidays("US", columns=["y"])
mp.fit(self.df)
future_df = mp.make_future_dataframe(7)
y_model = mp.model_pool["y"]
self.assertEqual(14, len(y_model.prophet.train_holiday_names))
y1_model = mp.model_pool["y1"]
self.assertIsNone(y1_model.prophet.train_holiday_names)
def test_cross_validation(self):
mp = multi_prophet.MultiProphet(columns=PREDICTOR_COLUMNS)
mp.fit(self.df)
cross_validation_dfs = mp.cross_validation(horizon="365 days")
for c, cross_validation_df in cross_validation_dfs.items():
self.assertTrue("ds" in cross_validation_df.columns)
self.assertTrue("yhat" in cross_validation_df.columns)
self.assertTrue("yhat_lower" in cross_validation_df.columns)
self.assertTrue("yhat_upper" in cross_validation_df.columns)
self.assertTrue("y" in cross_validation_df.columns)
self.assertTrue("cutoff" in cross_validation_df.columns)
def test_performance_metrics(self):
mp = multi_prophet.MultiProphet(columns=PREDICTOR_COLUMNS)
mp.fit(self.df)
performance_metrics_dfs = mp.performance_metrics(horizon="365 days")
for c, performance_metrics_df in performance_metrics_dfs.items():
self.assertTrue("horizon" in performance_metrics_df.columns)
self.assertTrue("mse" in performance_metrics_df.columns)
self.assertTrue("rmse" in performance_metrics_df.columns)
self.assertTrue("mae" in performance_metrics_df.columns)
self.assertTrue("mape" in performance_metrics_df.columns)
self.assertTrue("mdape" in performance_metrics_df.columns)
self.assertTrue("coverage" in performance_metrics_df.columns)
def test_add_regressor_single_model(self):
mp = multi_prophet.MultiProphet(columns=PREDICTOR_COLUMNS)
mp.add_regressor("Matchday", columns=["y"])
y_model = mp.model_pool["y"]
y_extra_regressor = y_model.prophet.extra_regressors["Matchday"]
self.assertEqual(10.0, y_extra_regressor["prior_scale"])
self.assertEqual("auto", y_extra_regressor["standardize"])
self.assertEqual(0.0, y_extra_regressor["mu"])
self.assertEqual(1.0, y_extra_regressor["std"])
self.assertEqual("additive", y_extra_regressor["mode"])
y1_model = mp.model_pool["y1"]
self.assertTrue(
"Matchday" not in y1_model.prophet.extra_regressors.keys())
def test_make_future_dataframe_dates(self):
dates = [
"2016-01-21", "2016-01-22", "2016-01-23", "2016-01-24",
"2016-01-25", "2016-01-26", "2016-01-27"
]
mp = multi_prophet.MultiProphet(columns=PREDICTOR_COLUMNS)
mp.fit(self.df)
future_df = mp.make_future_dataframe(7)
future_dates = [
np.datetime_as_string(d, unit="D")
for d in future_df.tail(7)["ds"].values
]
self.assertEqual(dates, future_dates)
def test_add_seasonality_single_model(self):
mp = multi_prophet.MultiProphet(columns=PREDICTOR_COLUMNS)
mp.add_seasonality(name="monthly",
columns=["y"],
period=30.5,
fourier_order=5)
y_model = mp.model_pool["y"]
y_seasonality = y_model.prophet.seasonalities["monthly"]
self.assertEqual(30.5, y_seasonality["period"])
self.assertEqual(5, y_seasonality["fourier_order"])
self.assertEqual(10, y_seasonality["prior_scale"])
self.assertEqual("additive", y_seasonality["mode"])
self.assertIsNone(y_seasonality["condition_name"])
y1_model = mp.model_pool["y1"]
self.assertTrue("monthly" not in y1_model.prophet.seasonalities.keys())
def test_constructor(self):
mp = multi_prophet.MultiProphet(columns=PREDICTOR_COLUMNS)
self.assertIsNotNone(mp)
