Experiment (dataset) update process (life-cycle)?

RadovanOresky — Tue, 14 Feb 2023 06:52:14 GMT

Hi, is it possible to update the dataset which is used in the ML experiment?

I have training data that is updated every day and (possibly) new rows are added. What I would like to achieve is to use the same ML experiment that is already setup, and just see that the training dataset was updated, run the new version and deploy a better model.

But, when I look at the experiment's dataset overview, I see that it didn't update and only initial values are taken into training.

Or is this just not accurate and autoML will take new data into consideration when I run the next version of the experiment?

Re: Experiment (dataset) update process (life-cycle)?

RadovanOresky — Tue, 14 Mar 2023 12:17:30 GMT

I researched it from multiple angles and figured out that this is currently not possible. I also realized that it would be helpful to be able to update the available ML deployment by a new version of a model.

I summarized both suggestions in a new idea:

https://community.qlik.com/t5/Suggest-an-Idea/AutoML-Experiment-dataset-and-Deployment-update-process-life/idi-p/2048966

topic Experiment (dataset) update process (life-cycle)? in Qlik Predict

Experiment (dataset) update process (life-cycle)?

Re: Experiment (dataset) update process (life-cycle)?