I’m going to be the first to admit this is hardly the most sophisticated/best/right way to do this going forward, but this has been available to play with for less than a week, so cut me some slack here.
“I can finally improve that sarcasm detector model!”
For those of you who don’t know what Azure Machine Learning is, it is the new cloud-based predictive analytics service Microsoft announced and made available for preview this past week. I hate to call it “predictive analytics for the masses”, but it’s certainly widening the audience considerably to include both the hardcore data scientists and those newer to the idea of predictive analytics.
To get started, other than signing up for an account in Microsoft Azure (duh), the easiest way is to follow the steps in this tutorial -
Once you’ve finished the tutorial, go to the Experiment Items on the left-hand side of your Machine Learning workspace and drag over the “Writer” module listed under “Data Input and Output”
I deleted the “Evaluate Model” module and connected the “Writer” module to the “Score Model” module.
You have a couple different options where you can output the data – I chose SQL Azure. You have to already have the table created in the SQL Azure Database, and you have to make sure you have the items from the output lined up 1:1 for each SQL column.
One last caveat – if you published it as a web service, you need to delete that before you try running this, or it will fail. At least I had to anyways. Just hit run and voila – the results should be in your SQL Azure table for you.
From there, getting it into LightSwitch or Power Pivot is a snap!
That’s it – like I said, I am sure as we all get more familiar with the ins and outs of Azure Machine Learning, and start to really dig into the web API and batch operations, the methods you use to get at the results will be far more sophisticated. But at least now you can show off to your boss a little in the interim.
Here is a link to the Azure Machine Learning Blog to learn more -
Until next time!