cancel
Showing results for 
Search instead for 
Did you mean: 

Problems creating Predictive Model based on custom Scenario with HRF

chrigu1
Participant
0 Kudos

Hi everyone,

I am trying to get some hands-on creating a custom heuristic score with HRF on 1709SP01 on-premise. I would like to sketch out the steps I undertook so far (might be interesting for some). Unfortunately, when I create a Predictive Model and try to change its status to "In Preparation" I receive an error.

I created a simple table to serve as a basis for my HRF rocabulary and hana information model:

create table "GUENTHEC"."BP_SCORING_01"("CONTACT_KEY" VARBINARY(16), "BP_ID" NVARCHAR(10), "PROFITABILITY" INTEGER, "SCORE" DOUBLE );

insert into "GUENTHEC"."BP_SCORING_01" ("CONTACT_KEY", "BP_ID", "PROFITABILITY") values ('020A778B7DFB1EE78989A596AAF78B80', '1234567890', 5000);

Then I created a information model (simple calculation view)

And based on it a simple vocabulary (notice I renamed the data object to "SCORE")

Then I moved on to the customizing.

First I created a Segmentation Object where I assigned the data source SAP_CE_INTERACTION_CONTACT.

Then I created a new data source from my previously created information model and assigned it to the Segmentation Object:

Afterwards, I created (and released) a Target Group from this profile containing the member we inserted in the table at the beginning.

Now to the customizing of the Predictive Scenario:

First I created a new Predictive Data Source which points to the vocabulary created above:

Next, I created an Implementation Method:

Then I created a new Application Scope:

Eventually, I put all together by creating a new predictive scenario:

As a final step I am trying to create the Predictive Model and that's where I am facing an issue where I am stuck:

On clicking the "In Preparation" button on the bottom right I receive the following message:

A model cannot be set to In Preparation when based on a training set that contains no responses regarding the target variable

The model does not move into the preparation stage so that I could create a model fit for it.

If anybody has a hint what I am doing wrong, it would be highly appreciated! I am pretty much stuck right now.

Thanks Christopher.

Accepted Solutions (0)

Answers (2)

Answers (2)

carsten_heuer1
Participant
0 Kudos

Hello Christopher,

it is difficult to investigate the issue from this problem description. Could you please open a ticket on component "CEC-MKT-ML-PRE" and give system access so that we can investigate this issue?

Many thanks and best regards,

Carsten

chrigu1
Participant
0 Kudos

Hi Carsten,

did that just now.

Thanks.

former_member197728
Active Participant
0 Kudos

Hey Christopher,

This might be an easy one. Can you try with setting the score to "1" instead of "5000" in your test data set?

In the predictive scenarios the expected score value in the training data set is usually the binary "1" or "0".

Best regards,

Tim

chrigu1
Participant
0 Kudos

Hi Tim,

thank you for your reply!

I did the following:

update "BP_SCORING_01" set "PROFITABILITY" = 1, "SCORE" = 1 where "CONTACT_KEY" = '020A778B7DFB1EE78989A596AAF78B80';

So I set "SCORE" to 1 and I changed "PROFITABILITY" from 5000 to 1.

But I still get the same result.

I actually don't understand why I have to provide a Training Set in the first place. This scenario is based on a heuristic model. I want to define my model fit in form of a decision table were I define e.g. if PROFITABILITY > 2500 then set SCORE = 0,75.

But all my attempts to be able to set the predictive model "in preparation" without the input of a training set failed so far.

Regards Christopher.

former_member197728
Active Participant
0 Kudos

Hello Christopher,

sorry, actually read "predictive" too many times and missed the "heuristic" in the first sentence of the question..

I have forwarded your question to the product colleagues.

Best regards,

Tim

chrigu1
Participant
0 Kudos

Hi Tim,

sorry, should maybe have pointed that out more clearly. Thanks a lot for passing on my issue.

Regards Christopher.

chrigu1
Participant
0 Kudos

Hi tim.nusch,

is there any news from your product colleagues on that matter or should I create an incident for my problem?

Regards Christopher.