Taylor G Smith

icon indicating a website link http://alkaline-ml.com icon indicating an email [email protected]

icon company @toyota-connected icon location Dallas, TX icon location Haskell is lame and you know it, Wheeler

Results 65 comments of Taylor G Smith

How can I reduce memory usage when fit auto arima ?

I think some of the recent changes in #359 and #361 might help with this. Hoping to get v1.7.0 out in the near future.

HEGY test of seasonality

Still would like to see this get added at some point 😬 Will try to find time soon..

support holes in the series with fb_prophet style fourier terms

Hmm this is pretty interesting. Could you by chance share a data snippet for me to tinker with? I like the idea of this a lot

Restricting number of BLAS threads makes ARIMA faster

Thanks for this very well-documented request. This is certainly something we can add to the doc for future users who stumble onto similar issues

Add more scoring to pmdarima.arima.AutoARIMA

Thanks for the willingness to contribute the code! This would be a good feature

checkresiduals equivalent

Cool, thanks @christopher-siewert . We should be able to access those methods through the results wrapper we get back from the fitted statsmodels model ```python model.arima_res_.model # the original model...

Support `simulate` after fitting with `auto_arima`

Thanks for the issue! This seems very reasonable. Would you be interested in contributing the feature? I agree calling with the `arima_res_` is not ideal, but it's a good work-around...

auto_arima returns flat predictions for near-zero scaled time series

Hi @claudia-hm sorry for the late reply. I'm looking into this. First thing I notice is a peculiar statsmodels warning when I enable the trace on the fit: ```python In...

diff_inv is not using xi parameter

Thank you for pointing this out. We will have this fixed in the next release.

Stepwise search with AICc fails in autoarima

Could you please provide a sample of data that reproduces this issue?