Microsoft Launches Azure Machine Learning Service, Developers Getting On Board With Predictive Analytics

Microsoft launched a new Azure cloud service Thursday which packages up the machine learning technology that powers many of its popular products and makes it available to developers.

The service, called Azure Machine Learning (ML), is accessible through any Web browser and features a drag-and-drop interface that's easy for developers and data scientists to use, Scott Guthrie, executive vice president of Microsoft's Cloud and Enterprise group, said in a blog post Thursday.

Since launching the Azure ML preview last July, Microsoft has added support for the Python and R programming languages along with several other new features, Guthrie said in the blog post.

"With today's General Availability release you can easily discover and create Web services, train/retrain your models through APIs, manage endpoints and scale Web services on a per-customer basis, and configure diagnostics for service monitoring and debugging," said Guthrie in the blog post.

Machine learning, a term that describes how computers can improve their functioning by ingesting and analyzing data, is getting lots of attention from Microsoft developers. Many are building apps that use predictive analytics, a form of machine learning that crunches historical data to predict what will happen in the future.

[Related: Sources Say Microsoft To Support VMware Workloads On Azure Public Cloud Starting In Q3]

Kristian Kimbro Rickard, CEO of MAX451, a Seattle-based Microsoft partner that's part service provider, part ISV, said Azure ML makes sense to customers once they understand how it can help solve business challenges.

That's one role that MAX451 is playing in order to get its customers familiar with what machine learning can achieve, she said.

"You really can't have a conversation with a chief marketing officer about machine learning," she said. "They know they want predictive analytics and they want to know their customers better, but it starts with understanding the business problem."

Chris Matty, CEO of Versium, a Seattle-based Microsoft partner that uses Azure ML in an app that helps retailers with fraud prevention, told CRN the service is helping his firm deliver more accurate fraud scores.

Since Azure ML runs on Microsoft's public cloud, Versium can add more external data to its fraud modeling system than they'd be able to with an on-premise machine learning setup, he said.

All this extra data has enabled Versium to fine-tune its fraud detection system and dramatically reduce false positives, said Matty. "Being able to pull in external data significantly improves predictive accuracy," he said.

Machine learning software has traditionally been expensive to use and difficult to manage, but Microsoft believes its pricing will make it more accessible to developers.

Microsoft is offering a free version of Azure ML that includes the ML Studio Service and lets users work with up to 10 GB of their own data.

There's also a paid version, called Standard, which will cost 38 cents per hour for the ML Studio Service and 75 cents per hour for the ML API Service until March 31.

After that, Microsoft will charge $9.99 per seat monthly, plus $1 per hour for Studio experiments and $2 per hour for production API compute.

Microsoft has been developing machine learning technology since 1992 and uses it today in Xbox, Bing and its Cortana digital assistant for Windows Phone, among other products.

But now that Azure ML is on the market, it's likely that Microsoft developers will take the technology in all sorts of new and interesting directions.

"Machine Learning and predictive analytics will fundamentally change the way all applications are built in the future," Guthrie said in the blog post.

PUBLISHED FEB. 19, 2015