ValueError for backtesting_forecaster when interval is provided #177
Comments
|
As far as I understood, the error comes from this file: you can see folowing lines: predictions_interval = self._estimate_boot_interval(
steps = steps,
last_window = copy(last_window_values_original),
exog = copy(exog_values_original),
interval = interval,
n_boot = n_boot,
random_state = random_state,
in_sample_residuals = in_sample_residuals
)
predictions = np.column_stack((predictions, predictions_interval))
predictions = pd.DataFrame(
data = predictions,
index = expand_index(
index = last_window_index,
steps = steps
),
columns = ['pred', 'lower_bound', 'upper_bound']
)according to your documentation: """
Returns
-------
prediction_interval : numpy ndarray, shape (steps, 2)
Interval estimated for each prediction by bootstrapping:
first column = lower bound of the interval.
second column= upper bound interval of the interval.
"""
I think Do you agree with that or is their something I don't understand? |
|
Hi @tkaraouzene, @JavierEscobarOrtiz I think we should explain this better in the docs. |
|
Hi ! Thanks for this information. def check_interval(interval: Optional[list[int]]) -> None:
"""Check provided confidence interval sequence is valid.
Parameters
----------
interval : list[int], optional
Confidence of the prediction interval estimated. Sequence of percentiles
to compute, which must be between 0 and 100 inclusive. If `None`, no check are made.
"""
if interval is not None:
if len(interval) != 2:
raise ValueError(
f"Interval sequence should contain exactly 2 values, respectively lower and upper interval bounds."
)
lower_bound = interval[0]
upper_bound = interval[1]
if lower_bound >= interval[1]:
raise ValueError(
f"Lower interval bound ({lower_bound}) has to be strictly smaller than upper interval bound "
f"({upper_bound}."
)
if interval[0] < 0:
raise ValueError(f"Lower interval bound ({lower_bound}) has to be >= 0.")
if interval[1] > 100:
raise ValueError(f"Upper interval bound ({upper_bound}) has to be <= 100.") |
|
I like this check @tkaraouzene We will add it in the new release. |
JavierEscobarOrtiz
added
enhancement
|
Hi @JoaquinAmatRodrigo ! If you want I can propose a pull request |
|
Hi @tkaraouzene, |
|
Hi @JoaquinAmatRodrigo !
I didn't see you've added the code so I made a PR (I saw your code thanks to the merge conflicts) 🤦♂️...
Yes perfect, thank you |
|
Hello @tkaraouzene, We are closing this issue. Thank you very much for your help. You can see it mentioned in the changelog: [0.5.0] - [Dev]Added
... |
Hi !
I'm trying to use your
backtesting_forecasterand when I use and ask for intervals, it leads to
ValueErrorWhen no intervals are asked, all works perfectly:
[in]:
[out]:
however asking for intervals leads to
ValueError[in]:
[out]:
The text was updated successfully, but these errors were encountered: