search

antoinecarme / pyaf

PyAF is an Open Source Python library for Automatic Time Series Forecasting built on top of popular pydata modules.
BSD 3-Clause "New" or "Revised" License
458 stars 73 forks source link

How to forecast future values using Croston Method? #109

Closed KathyGCY closed 5 years ago

KathyGCY commented 5 years ago

Hi @antoinecarme #, I'm replicating the examples provided in this notebook croston And It appears we only have fitted values (ending on 2002-12-31 ). What if I want to forecast future values? Say I want 7-day-out forecast?

In [In. 16] we see "df2['forecast'] = coeff * df2['q_est'] / df2['a_est']" So the forecast is calculated using the q_est and a_est.

Does that mean to forecast, I would do forecast 7-day-out q_est and a_est first? And after I arima/ets/bsts/other-TS forecast the q_enst and a_est, I then use the formula above to get the 7-day-out forecast? @

antoinecarme commented 5 years ago

Hi @KathyGCY,

Thanks for using PyAF.

This notebook is a notebook I used for prototyping (before Croston method is implemented). I keep track of everything .sorry.

Croston method can be used like any other model, You can find some python examples here :

https://github.com/antoinecarme/pyaf/tree/master/tests/croston

Your feedback is welcome.

antoinecarme commented 5 years ago

Added a README file for warning about the notebooks_sandbox directory :

https://github.com/antoinecarme/pyaf/blob/master/notebooks_sandbox/README.md

anuliketat commented 5 years ago

Hi @antoinecarme I'm trying to replicate R's Croston function in Python. I need some help in building that function from scratch. Also PyAFs auto forecasting engine is taking too long to train. Is it a bug? How to overcome it? Thanks.

antoinecarme commented 5 years ago

@anuliketat

Could you please create a separate issue and give more feedback on the problems you have , a dataset, a script etc.

antoinecarme commented 5 years ago

Closing issue after no response for 30 days. Not blocking.

Please repoen if needed.