Need specifics on how to setup a scheduler to kick off an experiment

Question

I need to kick off an experiment that includes a reader that pulls data from azure sql and creates a web service. I need kick off an experiment nightly and have a web service that pulls the data that got scored nightly. I have an exact experiment that can be seen http://oliviaklose.com/machine-learning-9-content-based-filtering/. What are the exact steps that I need to modify this experiment that runs around 12 min and what how do I set up a nightly scheduler that kicks it off so that I can use a published web service in my app? What is the easiest path? I do see BES url in azure ml studio and I also see folks respond using ADF but I don't actually see how to do what I'm asking. Thanks.

Answer 1

Hi JimJohnson,

Have you had a chance to review this document:

https://azure.microsoft.com/en-us/documentation/articles/data-factory-azure-ml-batch-execution-activity/

For specifics on Data Factory, pls use the Azure Data Factory forum: https://social.msdn.microsoft.com/Forums/en-US/home?forum=AzureDataFactory&filter=alltypes&sort=lastpostdesc

Regards,
Jaya

Answer 2

Yes, I did see that. Thank you. However, it's so complex, I've read it twice, I still don't understand how to schedule. If that's the path, I'll move to ADF forum.

Answer 3

Hi JimJohnson,

If you have a web service that is published and needs to be invoked on a regular basis, Azure Data Factory might be a good option.

Are you trying to re-create a new experiment automatically or simply invoke a published web service on a schedule?

Regards,
Jaya

Answer 4

Sounds like you want to retrain your model every night with new training data, then score some data with the web service that contains the newly trained model. Is that right?

If so, have you looked at this?

https://azure.microsoft.com/en-us/documentation/articles/machine-learning-retrain-models-programmatically/