Monday, August 29, 2016

Azure Machine Learning: Getting Started

This week, we're going to start talking about Microsoft's new Cloud Data Science offering, Azure Machine Learning.  Let's start with a little terminology.  What we're referring to as Data Science, also known as Predictive Analytics, Data Mining and Machine Learning, is the task of trying to get algorithms to make complex decisions based on data.  It's a huge area that we are super excited about.  For the past decade or so, the only major Microsoft Data Science offering in the Microsoft BI world was Data Mining in SQL Server Analysis Services.

At every SQL conference, there would be a couple of speakers talking about Data Mining in SSAS.  However, there wasn't much usage around the community.  We authored a 31-part blog series on utilizing this functionality within Excel.  This was great for a learning exercise, but didn't have much impact.  We think there are a couple of reasons for this.

First, the tools in SSAS were not nearly as developed as those from other major vendors, like SAS.  Second, we never encountered encountered any businessmen who were ready for the kind of "uncertainty" that comes from predictive modeling.  They wanted to know last month's sales compared to this month's inventory, and so on.  Well, it's starting to look like times are changing.  More and more businessmen seem to be getting more comfortable (and educated) when it comes to Data Science, and this is great for us.

This is where Azure ML comes in.  It's Microsoft's "new kid on the block" and it's far superior to the old SSAS functionality.  For this post, we won't dive into too many specifics.  But, you can immediately tell that we're not in Kansas anymore.
Sample 1 Workflow
For those of you that have used SSIS or some other ETL tool with a GUI, you'll recognize this type of layout.  But we have to say, it definitely looks better than anything we've used before.

So, how do you get started?  It's easy, just go to the Azure Portal and select New -> Intelligence -> Machine Learning.
Create Machine Learning Environment
From here, all you have to do is decide on names and all that boring stuff.  We will comment that Azure Machine Learning and SQL Azure are NOT free.  However, if you're just tinkering around on the weakest machines doing small stuff, you'll probably get by on a couple bucks per month.  There is great news though!  You may be able to get free Azure credits (read: money) on a one-time or recurring basis.  So, check with your IT team to see if your organization has access to these types of subscriptions.

In the coming posts, we're going to start looking through the free samples to see what sort of cool things this tool can do.  Thanks for reading.  We hope you found this informative.

Brad Llewellyn
BI Engineer
Valorem Consulting

No comments:

Post a Comment