Azure Machine Learning Now Available as a Function in Azure Stream Analytics

posted in: Uncategorized | 0

by Dann Anthony Maurno, Assistant Editor

By popular demand, Azure customers can now apply Azure Machine Learning (ML) models as a function on top of streaming data to get real-time insights. So wrote Microsoft Senior PM Sudhesh Suresh in a blog post about this preview.

Azure ML is largely a recommendation engine – an enterprise-grade version of, for example, the Netflix recommendation engine that tells you “Because you liked ‘Jaws,’ you might like…”. Both Dynamics CRM 2016 and Dynamics AX 7 (developed with the latest Azure capabilities) incorporate Azure ML for advanced analytics and decision making.

According to Suresh, numerous Azure customers have asked to combine the real-time analytics capabilities in Azure Stream Analytics with the power of Azure Machine Learning (ML) in quickly building and operationalizing any machine learning model as a web service.

This Azure ML/Azure Stream Analytics capability is now in public preview, and enables you to “score” individual events of streaming data, leveraging an ML model hosted in Azure. This enables you to, for example, build an application for real-time Twitter sentiment analytics. Two more real-world scenarios the Azure ML team is implementing with customers:

  • Using the capability to provide real-time product recommendations on the company website, helping them drive more revenue. Recommendations are served in real time based on website click data, user profile and other contextual information that is scored against an Azure ML product recommendation model.
  • Extracting in real time the topics and sentiment associated with conversations between customers and support staff. Support managers use this information to become aware of any critical customer issues in a timely manner, which in turn improves customer satisfaction and retention.

As those examples underscore, the most obvious utilities are largely in customer relationship, sales and marketing automation – hence Azure ML’s utility in Dynamics CRM. As we reported earlier this month about the general availability of Dynamics CRM 2016, Jujhar Singh, the new general manager of Dynamics CRM, said:

“We’re bringing all Microsoft has to offer in productivity and intelligence into a single experience…We’re bringing the advanced analytics and machine learning capabilities of the Cortana Analytics Suite to preview our first intelligent, adaptive processes for sales, customer service and social.”

According to Singh, the integration with Cortana will enable “intelligent selling” with cross-sell recommendations so sales reps can predict which products and services customers will need at various stages of the sales cycle.

But of course, advanced decision-making is needed enterprise wide, hence the Azure ML capabilities of the new Dynamics AX; also the ability to pluck those insights directly from Azure Stream Analytics.

Getting Started

This feature is available in the Azure portal, under the Azure Stream Analytics service. There you will find a new option called FUNCTIONS (see graphic) which enables you to add the Azure ML web service as a function. The ability to get ML scores by aggregating multiple events is not yet supported but is in the works.

The Microsoft Azure team provides a  getting-started tutorial  designed to help you quickly set up a simple Stream Analytics job with Machine Learning integration.

About Dann Anthony Maurno

Dann Anthony Maurno is a seasoned business journalist who began his career as International Marketing Manager with Lilly Software, then moved on as a freelancer to write for such prestigious clients as CFO Magazine; Compliance Week;Manufacturing Business Technology; Decision Resources, Inc.; The Economist Intelligence Unit; and corporate clients such as Iron Mountain, Microsoft and SAP. He is the co-author of Thin Air: How Wireless Technology Supports Lean Initiatives(CRC/Productivity Press, 2010).

Dann can be reached at