search

alkaline-ml / pmdarima

A statistical library designed to fill the void in Python's time series analysis capabilities, including the equivalent of R's auto.arima function.
https://www.alkaline-ml.com/pmdarima
MIT License
1.57k stars 231 forks source link

Support `statsmodels.tsa.arima.model.ARIMA` #526

Open jasmineliaw opened 1 year ago

jasmineliaw commented 1 year ago

Describe the question you have

Hi, may I ask is there any reason that statsmodels.tsa.arima.model.ARIMA is not supported in pmdarima?

This post from statsmodels mentioned that:

Finally, there is a new model class ARIMA, which is meant to (eventually) be a single point of entry for all ARIMA-type models (including SARIMAX models). It is a subclass of statespace.SARIMAX but allows fitting using any of the estimators above.

And imo ARIMA could be more performant than SARIMAX because the different methods (estimators) provided in ARIMA.fit() could potentially speed up the fit process (see the inline comment in statsmodels.tsa.arima.model.ARIMA).

Reference of ARIMA's fit method: https://www.statsmodels.org/devel/generated/statsmodels.tsa.arima.model.ARIMA.fit.html#statsmodels.tsa.arima.model.ARIMA.fit

So would you consider adding / replacing the current implementation with statsmodels.tsa.arima.model.ARIMA?

Please let me know if you have a different opinion.

Thanks!

Versions (if necessary)

No response