Hi SAP PA team,

Firstly I would like to thank you all for your continued efforts in area of Machine learning/PA and making it more and more automated.

However due to so many products being around: SAP PAL, APL, BO - PA, I would be really happy if someone could clarify these simple queries we have and are facing a hard time going with the best approach.

Consider I have a situation where I have to implement time series forecasting but as per the data it could either be ARIMAX or CROSTON because real data could be anything but as a human being I can decide as per the data or even write some code rules to detect periodicity and switch automatically between ARIMAX or CROSTON.

Questions:

1. SAP BO-PA uses Auto.KXEN models to find the best time series model. Is it smart enough to determine by itself that a particular set of data is suitable for CROSTON or ARIMAX ? Would you prefer going with Manual or Auto model in this case ?

2. If I don't have an option of using SAP BO-PA, can I use SAP APL wrappers as a replacement of BO-PA? Is the result of "Forecast" function of APL and BO-PA going to be same?

3. If BO-PA is same as APL and I am using SAP APL as replacement of SAP BO-PA, there are scenarios when I have to call these wrappers from ABAP. What is the best way to call it through ABAP ? I used to create wrapper functions and table types of PAL algos manually and then create an AMDP class to call these wrappers later on. Any example blog/doc will be really helpful ?

ex: call "_SYS_BIC"."PAL_POLYNOMIALR_PROC"( p1 => :input_table, p2 => :control_table, p3 => :result_table, p4 => :fitted_table, p5 => :signif_table, p6 => :model_table );

4. If in case SAP BO-PA is not same as APL then, how can I call the SAP BO-PA algorithms from ABAP stack ?

Thanks,

Hasan