Dear community,
i'm facing the following issue.
A customer wants to pre-categorize e-mails that flow into the dataverse based on the text they contain (spam, newsletter, ham). For this i created a AI Builder category classification model, which actually works fine on the classficication part. Since the customer wants the model to get better and better every week, the model needs to be retrained. The process works fine, if I do that every week by hand, but that is work every week that the customer does not want to pay for the next serveral years.
There is a beta feature for prediction datasets named "shedule" but this is not availible for classification datasets as a support ticket told me.
This is an absolut critical point for my customer. He does NOT want to use AI Builder, if its not able to retrain every week by shedule.
Do I have any possibility to train the model by shedule? Every information helps. I'm fine, if its a flow, a Runbook or whatever. I'm familiar with all the automation technologies I can use in the microsoft / dataverse technology stack. I just need this to retrain somehow.
Can somebody please help me out here?
How is the data currently getting pulled for you to retrain the model? If all the miscategorized emails can be aggregated in an CSV or a sharepoint list, I'm sure you could use RPA or a flow to have that uploaded into your model.