Skip to Content
0
Aug 10, 2021 at 12:51 AM

Questions about SAP PA Time Series

170 Views Last edit Aug 10, 2021 at 02:25 AM 8 rev

Dear Everyone:

I am doing a project using Predictive Analytics Time Series.

My questions are

1. In "Apply Model", the applying dates should belong to the training period.

Can I apply data with dates not belonging to the training period, to the model I generated?

If I cannot, what is the use case of "Apply Model"?

2. PA Time Series will not run if the last training date is the last data with the last date.

There will be error message "No future values detected after the last training date given"

Why should I select the last training date before the last date? I want to know the reason

3. In "Cutting Strategy", I can choose "Sequential", not "Sequential without test"

If I choose "Sequential", there will be test data among the total data input.

But I cannot find out which data belong to test data

So if choose "Sequential", which part in total data will belong to "training data" and which part will belong to "validation data"

and which part will belong to "test data"? I want to know exactly

4. Which theory does PA Time Series use to generate "forecasts"?

5. Can I add any explanatory variables to "Predictable Variables" list w.r.t the target variable?
That is, can I add any predictable variables just like doing multiple linear regressions?

6. If I change the "number of forecasts" in "Summary of Modeling Parameters" panel, will the model (which is generated) change?

7. PA Time Series will separate date set into "training data" and "validation data"

Then, do the data after the last training date not belong to either "training data" or "validation data"?

8. I want to know how many types of regressions PA Time Series will run if I choose "default" option in "Summary of Model Parameters" --> "advanced" --> "Modeling Procedure"

Thank you and have a nice day