Questions about SAP PA Time Series and ADS(analytical dataset)

[former_member666992]

Dear everyone:

My client is using SAP PA and they are going to use Time Series.

I have two questions.

1. The dataset will have several extra-predictable variables besides the target variable.

I want to predict the targets for several days after the last training date without giving extra-predictables for those several days (after the last training date)

Is this possible using PA Time Series?

2. In the same situation as in the question 1,

if I use ADS method (analytical dataset)

can I predict the targets for several days after the last training date without giving extra-predictables?

Thank you in advance and have a nice day!!

[marc_daniau · Accepted solution]

The extra predictors values are required in the past and in the forecast horizon, regardless of the data preparation tool you are using: ADS, SQL script, etc.

[former_member666992 · Accepted solution]

Thank you very much, Marc.

Now I understand fairly all right.

Have a nice weekend and see you again.

Take care

[marc_daniau]

Extra Predictors in PA Time Series work that way.

You have a column in your daily series, say: Is_holiday (Yes/No).

Let's assume PA learns from the historical data that this column is helpful to make more accurate forecast compared to a model using only the date. The holiday information in the forecast horizon (after the last training date) is required to apply the trained model.

If for some reason you cannot provide the extra predictor values in the future, don't use it ; the model then will extrapolate the series based on the date information only.

[former_member666992]

Dear Marc:

Thank you for your comment.

Then, as I asked you in the main content, can I predict future targets even without future extra-predictor values of future

if I use ADS technique ?