auto-arima in arima model

[vinitpathri]

i am creating a model for the prediction of license usage in our environment.

tried many combination(around 25) of p,d,q for ARIMA, got few results which are best among the rest .

i want to know is it possible to use auto-arima in splunk?
if yes the please share the SPL.

Thanks in advance

[grana_splunk]

auto Arima is not there in MLTK today but I would suggest to look into the new statespace algorithm provided out of box in MLTK.

[vinitpathri]

as in Kalman filter?

[vinitpathri]

actually not able to do much experiments with kalman filter and with default parameters i am not getting satisfactory results.

[grana_splunk]

What parameters you're looking for? you can check the documentation here : https://docs.splunk.com/Documentation/MLApp/5.1.0/User/Algorithms#StateSpaceForecast

Also, Arima ( not auto arima) is available out of the box in MLTK. Check that option too.

[grana_splunk]

It is Kalman filter with an ability to save and reuse the model..what's your usecases?

[vinitpathri]

i want to predict the license usage in my environment.
so my dataset is simple with 2 fields,
"date" and "license used on that date".

[tauliang]

Auto-ARIMA is supported in other languages like R, however, I don't see a way to use it out of the box with Splunk MLTK.

Of course, one can always add a custom algorithm to the MLTK. I'd love to know as well how one can use auto-ARIMA without writing any custom code.