Microsoft Unveils Machine Learning Service for PredictionDina Bass
Microsoft Corp. is introducing a new Internet-based service that helps companies use data analysis to predict behavior, in a bid to gain more customers in the fast-growing field of machine learning.
The service lets users more easily build algorithms to predict purchasing patterns, electricity usage and more. The work and results can be hosted on the Web through Microsoft’s Azure cloud offering. The new product, dubbed Azure Machine Learning, has been in a trial period for about a year and now has about 100 customers testing it, said General Manager Eron Kelly.
Microsoft, under new Chief Executive Officer Satya Nadella, is trying to boost its focus on cloud software as more companies gravitate to hosting data and other operations on the Web. Nadella ran the cloud business before becoming CEO in February. The Redmond, Washington-based software maker wants to bolster usage of Azure as it tries to narrow the gap with market leader Amazon.com Inc.’s Web Services business.
Machine learning, which is an artificial intelligence field that involves constructing systems that automatically analyze and learn from data, has been on the rise. Even as companies collect more information, the challenge has been how to extract value from it. As a result, the market for analysis software is growing about 18 percent a year and will reach $6.5 billion by 2019, according to Transparency Market Research.
“One of the challenges is as you create more data -- data in and itself isn’t all that valuable,” Kelly said. “It’s the insights you can create from it, and as the datasets get larger, it becomes harder to analyze.”
Microsoft said its new product can be used by full-fledged data scientists to speed up their work, as well as by analysts who aren’t trained in the mathematical underpinnings of the field.
Microsoft’s own retail stores are using the software to predict fraud, while another unnamed retailer is analyzing past customer purchases to decide future inventory, said Kelly. Carnegie Mellon University is tracking weather patterns and electricity usage in its buildings and has told Microsoft the system will enable them to save 30 percent on energy, he said.
Microsoft will offer a public preview of the service, which was code-named Project Passau, in July. It’s the product of work from the company’s research unit and its data platform team, helmed by Joseph Sirosh, a machine-leaning executive who Microsoft hired from Amazon in July 2013.
“Soon machine learning will help to drastically reduce wait times in emergency rooms, predict disease outbreaks and predict and prevent crime,” Sirosh wrote in a blog post. “To realize that future we need to make machine learning more accessible.”