"ARIMA

felix_w

Dear Rapidminer Community, 

 

I was reading into the ARIMA model and I understood that it is vital to have a look at the autocorrelation of the data in order to decide how to set p and q. My question is now, is autocorrelation somehow already implemented in the ARIMA operator and I missed it or how I can set my p,q and d?

I know I can do it via Optimize Parameters and I will get a result but other than that, setting p and q is just trial and error for me? 

 

Best regards

Felix

tftemme

Hi @felix_w,

 

Currently autocorrelation is not yet implemented as a time series operator. You can of course validate your ARIMA model either by its performance on the training data (described by aic, bic, aicc, see the performance outputport of the operator) or better by the performance on a test window (for example by using the Forecast Validation operator). 

 

Autocorrelation and partial autocorrelation functions will be added in the future.

 

Best regards,
Fabian