search

unit8co / darts

A python library for user-friendly forecasting and anomaly detection on time series.
https://unit8co.github.io/darts/
Apache License 2.0
7.92k stars 860 forks source link

Support for probabilistic predictions with historical_forecasts #1924

Closed ramiyaari closed 1 year ago

ramiyaari commented 1 year ago

I want to use the historical_forecasts method instead of calling fit and predict to evaluate the forecasting models over the whole timeseries. However, in this method there is no support for setting num_samples to obtain a probablistic prediction like in the predict method. So the timeseries returned by historical_forecasts doesn't support calling quantile_df like the timeseries returned by predict. Is that a missing feature by the package or is there a way to overcome this limitation? Is there a reason for this limitation?

madtoinou commented 1 year ago

Hi, which version of darts are you using?

In the latest version (0.24.0 at the time of writing), historical_forecasts() does have an argument num_samples (doc) that allows you to obtain a probabilistic forecasts.

ramiyaari commented 1 year ago

Thank you. Apparently, either I was looking at an older version or I missed it somehow.