Forecaster

ForecastKnownValues

Categorical in XInsamplePred int insampleMultivariate

Forecaster that plays back known or prescribed values as forecasts.

Takes a data set of “known future values” to produces these in the sktime interface.

Common use cases for this forecaster:

as a dummy or naive forecaster with a known baseline expectation
as a forecaster with (non-naive) expert forecasts, “known” values as per expert
as a counterfactual in benchmarking experiments, “what if we knew the truth”
to pass forecast data values in a composite used for postprocessing, e.g., in combination with ReconcilerForecaster for an isolated reconciliation step

When forecasting, uses pandas.DataFrame.reindex under the hood to obtain predicted values from y_known. Parameters other than y_known are directly passed on to pandas.DataFrame.reindex.

Schnellstart

python

from sktime.forecasting.dummy import ForecastKnownValues

estimator = ForecastKnownValues(y_known, method=None, fill_value=None, limit=None)

Parameter(4)

y_knownpd.DataFrame or pd.Series in one of the sktime compatible data formats: should contain known values that the forecaster will replay in predict can also be in a non-pandas sktime data format, will then be coerced to pandas
methodstr or None, optional, default=None: one of {None, ‘backfill’/’bfill’, ‘pad’/’ffill’, ‘nearest’} method to use for imputing indices at which forecasts are unavailable in y_known
fill_valuescalar, optional, default=np.NaN: value to use for any missing values (e.g., if method is None)
limitint, optional, default=None=infinite: maximum number of consecutive elements to bfill/ffill if method=bfill/ffill

Beispiele

>>> import pandas as pd
>>> y_known = pd. DataFrame (range (100))
>>> y_train = y_known [: 24 ]
>>> 
>>> from sktime.forecasting.dummy import ForecastKnownValues
>>> 
>>> fcst = ForecastKnownValues (y_known)
>>> fcst. fit (y_train, fh = [1, 2, 3 ]) ForecastKnownValues(
... ) The forecast “plays back” the known/prescribed values from y_known
>>> y_pred = fcst. predict ()