Hi Team, I wanted to create Einstein discovery(ED) Model for Client Churn. So i created CRMA recipe to gather details from 3+ objects.(Account,Task,ChildCustomObj1 etc.). So when I created model on such dataset(populated by recipe), ED Model is giving Prediction score for every row: (as shown in below table)
====Account|=====Task|=ChildObj1|Score
DemoCompany|DemoTask1|ChildRec11|31
DemoCompany|DemoTask2|ChildRec12|32
DemoCompany|DemoTask2|ChildRec13|33
DemoCompany|DemoTask2|ChildRec14|34
I wanted to calculate score on Account level but Einstein is calculating on every row. Actually i need to calculate score for Account 'DemoCompany' considering all it's child records.
All tasks: DemoTask1,DemoTask2
All ChildObj1 records: ChildRec11, ChildRec13, ChildRec13, ChildRec14
Is there way to give ParentChild records from CRMA to Einstein model, so i will get score on account level?
1 Answer
Unsure but I would try aggregating the child objects at the grain of the parents and then using the child object aggregates in the model to predict behavior of the parent
based on your headers above generate Account, unique(Task) over([..] partition by Account, unique(ChildObj1) over([..] partition by Account
if you can't run the prediction on grouped data then run it at the grain of the child objects and i would just sum the score in a partition and consider additional math like avg or dividing by row number
generate Account, unique(Task) over([..] partition by Account, unique(ChildObj1) over([..] partition by Account, sum(Score) over([..] partition by Account, avg(Score) over([..] partition by Account, (sum(Score) over([..] partition by Account / row_number() over([..] partition by Account)
totally taking a shot in the dark
- Your answer could be improved with additional supporting information. Please edit to add further details, such as citations or documentation, so that others can confirm that your answer is correct. You can find more information on how to write good answers in the help center.2024-10-23 16:47:27 +00:00Commented Oct 23, 2024 at 16:47