Azure Machine Learning democratizes analytics

  • April 09, 2015
  • Feature

By Bill Lydon, Editor

In August at the Microsoft Campus I had the opportunity to learn more about Microsoft’s Azure Machine Learning cloud-based predictive analytics software. It is designed to allow users to gain insights into their systems by leveraging data from those systems. I believe Azure is analogous to the power of Excel and Visual Basic - two products that democratized computing in the past. I personally tried Azure Machine Learning and was impressed by the ease of use.

I recently spoke with Microsoft’s Joseph Sirosh, Corporate Vice President of Machine Learning. Azure Machine Learning is part of Microsoft’s investment in a broad portfolio of new Azure solutions that empower users to leverage big data and cloud computing. Microsoft developed Azure Machine Learning to simplify the analysis of big data as part of their cloud and Internet of Things (IoT) strategy. Gaining insights from big data is a big part of gaining value from Internet of Things (IoT) investments. The industrial automation industry has been using data to gain insights for years. However, beyond creating alarms, higher level analysis has been expensive and difficult to program. Azure provides an economical way to use cloud-based historians coupled with powerful analytics to gain insights and improve operations. Microsoft’s goal is to bring big data into the mainstream.

Integrated Design Environment

Azure Machine Learning has an impressive drag and drop Integrated Design Environment (IDE). The Azure Machine Learning IDE is cloud-based and uses a web browser and therefore requires no other software. Users create applications by dragging and dropping functions and building data flow graphs to set up analytics. Machine Learning Studio features a library of sample experiments. Azure ML also supports R and Python custom code, which can be dropped directly into the workspace. R is an application-oriented programming language and software environment for statistical computing and graphics. The R language is widely used among statisticians and data miners for developing statistical software and data analysis. Azure Machine Learning models can be live within minutes as a fully managed web service.

Users drag and drop functions and create dataflow relationships by connecting functions.

Application Marketplace

Microsoft has created the Azure Marketplace where experts can share and sell functions they have created. The goal is to foster an ecosystem of industry experts and leverage knowledge.

Predictive & Preventative Maintenance Application

ThyssenKrupp Elevator worked with CGI, a Microsoft partner, to develop a predictive and preemptive maintenance program using Azure Machine Learning. Andreas Schierenbeck, CEO of ThyssenKrupp Elevator, said, “We wanted to go beyond the industry standard of preventative maintenance to offer predictive and even preemptive maintenance, thereby guaranteeing a higher uptime percentage on our elevators.”

Using Internet of Things (IoT) technology and concepts, data from sensors and systems at an elevator site is communicated to the cloud. Azure Machine Learning is used for predictive and preemptive maintenance analytics. The information is available to service people using the Internet and their mobile devices.

Watch the following video of Dr. Rory Smith, ThyssenKrupp Director of Strategic Development for Americas, describing the application.

Azure Event Hubs

Azure Event Hubs are a scalable publish/subscribe interface that can intake millions of events per second. The hubs allow users to process and analyze the massive amounts of data produced by connected devices and applications. They can be used with an onsite or cloud-based database.

Software as a Service (SaaS)

The offerings in the Microsoft Azure Machine Learning Service enable customers to pay only for the infrastructure they really need and spin up/down resources automatically based on actual usage. Current prices of the Standard Tier, effective April 1, 2015, are as follows:

  • ML seat subscription of $9.99/seat/month
  • Azure ML Studio usage: $1/studio experiment hour
  • ML API usage: Hourly-based - $2/production API compute hour OR Transaction-based - $0.50/1,000 production API transactions

More Azure Machine Learning Pricing.

You can give it a test drive for free. The Azure Machine Learning website offers learning resources, including videos.

Thoughts & Observations

Azure Machine Learning is an example of high-function, low-cost technology that was created to serve high-volume applications like IoT, big data, and analytics. The industrial automation industry is relatively small, and it has adopted technology such as PCs, Windows OS, and mobile computing when they are stable enough to provide more function at less cost.

OSIsoft is working with Carnegie Mellon to apply OSI PI and Azure Machine Learning to analyze complex and hard-to-predict building behaviors. Meet Carnegie Mellon’s energy sleuths.

Each year, I attend a closed-door Pharmaceutical Automation Roundtable. One of the member companies described how their IT people use off-the-shelf statistical software to analyze batch data. As a result, they found ways to improve yields that were not previously obvious. Tools like Azure Machine Learning will simplify this analysis.

I believe Azure Machine Learning is an example of things to come from IoT and big data developments, and one that can be applied in industrial applications. This is the tip of the iceberg.

Related Articles

Did you enjoy this great article?

Check out our free e-newsletters to read more great articles..