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
7.96k stars 866 forks source link

Can autoarima forecast under a confidence level? Thanks. #351

Closed PaulJXu closed 3 years ago

PaulJXu commented 3 years ago

Is your feature request related to a current problem? Please describe. A clear and concise description of what the problem is.

Describe proposed solution A clear and concise description of what the library should provide to solve missing functionality.

Describe potential alternatives A clear and concise description of any alternative solutions or existing features that might solve it.

Additional context Add any other context or screenshots about the feature request here.

pennfranc commented 3 years ago

Hi @jerpson, I think you're asking about whether you can produce confidence intervals as part of a prediction on top of point predictions, correct? Currently Darts does not support confidence intervals. However, we are planning to introduce this functionality as part of a big update where we will introduce probabilistic forecasting models. I am not yet sure whether this will include AutoARIMA, since we are wrapping a model from a different library, which as far as I can tell does not provide this functionality out of the box. But please stay tuned for our next updates!

hrzn commented 3 years ago

Hi @jerpson starting in v0.9.0 the model ARIMA is probabilistic - you just have to provide a num_samples argument to predict() in order to get back a probabilistic TimeSeries (for instance: model.predict(n=36, num_samples=1000), from which you can obtain quantiles for instance, by calling quantiles_df() or quantile_timeseries().

AutoARIMA is not (yet) probabilistic, but once you have figured out the best ARIMA parameters using AutoARIMA, you can fit a probabilistic ARIMA model.