def test_fit_regressor_visnights(load_df_and_hier_visnights):
hierarchical_vis_data, vis_hier = load_df_and_hier_visnights
ht = HTSRegressor(model="holt_winters", revision_method="OLS")
ht.fit(df=hierarchical_vis_data, nodes=vis_hier)
assert isinstance(ht.hts_result.errors, dict)
assert isinstance(ht.hts_result.models, dict)
assert "total" in ht.hts_result.models
def test_exog_fit_predict_fb_model(hierarchical_mv_data, mv_tree_empty):
exogenous = {
k: ["precipitation", "temp"]
for k in hierarchical_mv_data.columns
if k not in ["precipitation", "temp"]
}
horizon = 7
train = hierarchical_mv_data[:500]
test = hierarchical_mv_data[500:500 + horizon]
clf = HTSRegressor(model="prophet", revision_method="OLS", n_jobs=0)
model = clf.fit(train, mv_tree_empty, exogenous=exogenous)
preds = model.predict(exogenous_df=test)
assert isinstance(preds, pandas.DataFrame)
# base + steps ahead
assert len(preds) == len(train) + horizon
assert len(model.hts_result.forecasts["total"]) == len(train) + horizon
assert len(model.hts_result.errors) == len(train.columns) - 2
for column in train.columns:
if column not in ("precipitation", "temp"):
assert column in model.hts_result.errors
assert column in model.hts_result.forecasts
assert column in model.hts_result.residuals
def test_convenience(load_df_and_hier_uv):
hd, hier = load_df_and_hier_uv
hd = hd.head(200)
ht = HTSRegressor(model='holt_winters', revision_method='OLS')
ht = ht.fit(df=hd, nodes=hier)
preds = ht.predict(steps_ahead=10)
rev = revise_forecasts(
'OLS',
nodes=ht.nodes,
forecasts=ht.hts_result.forecasts,
errors=ht.hts_result.errors,
residuals=ht.hts_result.residuals,
)
assert isinstance(rev, pandas.DataFrame)
assert numpy.allclose(preds.values, rev.values)
rev_s = revise_forecasts(
'OLS',
summing_matrix=ht.sum_mat,
forecasts=ht.hts_result.forecasts,
errors=ht.hts_result.errors,
residuals=ht.hts_result.residuals,
)
assert isinstance(rev_s, pandas.DataFrame)
assert numpy.allclose(rev.values, rev_s.values)
rev = revise_forecasts('AHP',
nodes=ht.nodes,
forecasts=ht.hts_result.forecasts)
assert isinstance(rev, pandas.DataFrame)
def test_fit_regressor_uv(load_df_and_hier_uv):
hierarchical_sine_data, sine_hier = load_df_and_hier_uv
ht = HTSRegressor(model='holt_winters', revision_method='OLS')
ht.fit(df=hierarchical_sine_data, nodes=sine_hier)
assert isinstance(ht.hts_result.errors, dict)
assert isinstance(ht.hts_result.models, dict)
assert 'total' in ht.hts_result.models
def test_multiprocessing_fit(load_df_and_hier_uv):
distributor = MultiprocessingDistributor(n_workers=2)
hsd, hier = load_df_and_hier_uv
reg = HTSRegressor()
reg = reg.fit(df=hsd, nodes=hier, distributor=distributor)
node_keys = list(chain.from_iterable([v for v in hier.values()]))
for node in node_keys:
assert node in reg.hts_result.models.keys()
def test_dask_cluster_two_workers(load_df_and_hier_uv):
with Client(n_workers=1, processes=False) as client:
address = client.scheduler_info()['address']
distributor = ClusterDaskDistributor(address=address)
hsd, hier = load_df_and_hier_uv
reg = HTSRegressor()
reg = reg.fit(df=hsd, nodes=hier, distributor=distributor)
node_keys = list(chain.from_iterable([v for v in hier.values()]))
for node in node_keys:
assert node in reg.hts_result.models.keys()
def test_instantiate_revision(load_df_and_hier_visnights):
hierarchical_visnights_data, visnights_hier = load_df_and_hier_visnights
for method in ['OLS', 'FP', 'WLSS', 'WLSV', 'PHA', 'AHP', 'NONE']:
ht = HTSRegressor(model='holt_winters', revision_method=method)
ht.fit(df=hierarchical_visnights_data, nodes=visnights_hier)
rm = RevisionMethod(method,
sum_mat=ht.sum_mat,
transformer=ht.transform)
_ = ht.predict(steps_ahead=3)
revised = rm.revise(forecasts=ht.hts_result.forecasts,
mse=ht.hts_result.errors,
nodes=ht.nodes)
assert isinstance(revised, numpy.ndarray)
assert revised.shape == (11, len(ht.hts_result.forecasts))
def test_predict_regressor_visnights(load_df_and_hier_visnights):
hierarchical_vis_data, vis_hier = load_df_and_hier_visnights
hvd = hierarchical_vis_data
for model in ["holt_winters", "auto_arima"]:
ht = HTSRegressor(model=model, revision_method="OLS")
ht.fit(df=hvd, nodes=vis_hier)
preds = ht.predict(steps_ahead=4)
assert isinstance(preds, pandas.DataFrame)
# base + steps ahead
assert len(preds) == len(hvd) + 4
assert len(ht.hts_result.forecasts["total"]) == len(hvd) + 4
assert "total" in ht.hts_result.errors
assert "total" in ht.hts_result.forecasts
assert "total" in ht.hts_result.residuals
assert "NSW" in ht.hts_result.errors
assert "NSW" in ht.hts_result.forecasts
assert "NSW" in ht.hts_result.residuals
def test_predict_regressor_uv(load_df_and_hier_uv):
hierarchical_sine_data, sine_hier = load_df_and_hier_uv
hsd = hierarchical_sine_data.head(200)
for model in ["holt_winters", "auto_arima", "sarimax"]:
ht = HTSRegressor(model=model, revision_method="OLS")
ht.fit(df=hsd, nodes=sine_hier)
preds = ht.predict(steps_ahead=10)
assert isinstance(preds, pandas.DataFrame)
# base + steps ahead
assert len(preds) == len(hsd) + 10
assert len(ht.hts_result.forecasts["total"]) == len(hsd) + 10
assert "total" in ht.hts_result.errors
assert "total" in ht.hts_result.forecasts
assert "total" in ht.hts_result.residuals
assert "a" in ht.hts_result.errors
assert "a" in ht.hts_result.forecasts
assert "a" in ht.hts_result.residuals
def test_instantiate_regressor():
ht = HTSRegressor()
assert ht.model == "prophet"
assert isinstance(ht.hts_result, HTSResult)
assert isinstance(ht.hts_result.residuals, dict)
