Discover this year's submissions to the Dataiku Frontrunner Awards and give kudos to your favorite use cases and success stories!READ MORE

Clarity on the drift detection steps in the Production Quality Control Sample Project

razan
Level 1
Level 1
Clarity on the drift detection steps in the Production Quality Control Sample Project

Hello community, and good day.

I am trying to re-create the Production Quality Control Sample Project one of the sample projects in www.dataiku.com/learn/samples, In the drift detection and alerting steps the first step used is widow recipe to compute the average and standard deviation of the injection time per recipe. (As described in the projects Wiki and the knowledge base.)

flow_zone_drift_alerts.png

Here is the recipe settings:

Screenshot from 2022-09-20 16-46-57.png

Screenshot from 2022-09-21 08-36-03.png

And here is the output:

Screenshot from 2022-09-21 08-38-53.png

If I try it in my dss 11, the output is also empty. When I turned off the WINDOW FRAME option the output dataset InjectionTime_avg is equal to the InjectionTime and the InjectionTime_std is empty; since there is no window frame setting. Also if I tried the DateTime column as the ORDER COLUMN the output is also the same as if there is no window frame. 

Screenshot from 2022-09-21 09-03-34.png

Screenshot from 2022-09-21 09-06-08.png

 

Is there any other documentation (other than the WIKI and the knowledge base) to clarify how this step was done, or which column should be chosen in the ORDER COLUMN to obtain the drift detecting step correctly.

 

Thank you.


Operating system used: Ubuntu 22


Operating system used: Ubuntu 22

0 Kudos
0 Replies