Probabilistic Forecast Reconciliation

[Manjubn777]

After going through the package and paper found that supports only point forecasts reconciliation. I would like to know if it supports probabilistic forecast reconciliation. I appreciate your quick response to my inquiry for information.

[kdgutier]

Hi @Manjubn777,

We have been doing research/working on probabilistic hierarchical forecasting, but have not released.

I can point you to my work in progress paper "Probabilistic Hierarchical Forecasting with Deep Poisson Mixtures" and Taieb's PERMBU.

Regarding implementations of algorithms GluonTS team did an effort to improve the usability of PERMBU in this repository, it has a lot of dependency frictions.

If you have any pointers please let me know, this is a really interesting topic for us.

[AzulGarza]

Hey @Manjubn777! The new version of HierarchicalForecast supports probabilistic forecast reconciliation.

• Example assuming normality
• Example using bootstrapping

[Manjubn777]

Hey Team, Thank you so much for notebooks and I do have a question. Let's say I have 2000 forecast samples for each day,which will be passed as y_hat_df and I have actuals which is like one value for each day(daily data)., I will pass actual data as fitted values. setting bootstrap = True helps me reconcile?

I would appreciate your answer to the above question.

Thanks and Regards, Manjula

On Wed, Sep 28, 2022 at 7:39 AM fede @.***> wrote:

Hey @Manjubn777 https://github.com/Manjubn777! The new version of HierarchicalForecast supports probabilistic forecast reconciliation.

• Example assuming normality https://nixtla.github.io/hierarchicalforecast/examples/australiandomestictourism-intervals.htmll
• Example using bootstrapping https://nixtla.github.io/hierarchicalforecast/examples/australiandomestictourism-bootstraped-intervals.html

— Reply to this email directly, view it on GitHub https://github.com/Nixtla/hierarchicalforecast/issues/31#issuecomment-1260296842, or unsubscribe https://github.com/notifications/unsubscribe-auth/AQXBE72UYYZNT6FXOF4A2A3WAOSFDANCNFSM55ATCGMQ . You are receiving this because you were mentioned.Message ID: @.***>