Interview: Microsoft Azure Machine Learning set for July.
Microsoft is set to bolster its cloud portfolio with a predictive analytics tool hosted on Azure, boasting data-driven algorithms used in Bing and Xbox Live.
Azure Machine Learning will be available for a public preview in July, the tech firm announced as it promised the tool would cut analytics app development from weeks and days to hours and minutes.
The tech giant’s Data Platform general manager, Eron Kelly, told CBR the service was created to solve a problem for companies wanting to make accurate data-based predictions but who lacked the data scientists to do it.
He said: "It allows you to shift from analysing the past to anticipating the future. What’s really cool about this service is I literally push a button and I publish the model as a service. It’s running in the cloud and you don’t have to worry about hardware, health monitoring, or failover, this is all done for you.
"Because of the simple graphical interface, there’s a broader group of people who can use this service."
One of those is the company’s own Microsoft Stores, which have been using Azure Machine Learning to predict and cut future credit card fraud, using the service to build a model that tags real-time transactions as fraudulent or not based on historical data.
Kelly said the stores have seen up to a 20% reduction in fraud thanks to the model.
The tool lets customers build their own algorithms, but also includes "out-of-the-box" algorithms for Bing and Xbox Live, with Microsoft using the latter – which works out which games and films to promote on a gamer’s log-in screen based on their previous interests – to let retailers push products to customers when they visit their websites.
"We’ve been doing is using machine learning within Microsoft for a long time and as we build these algorithms we’ve made them available out of the box. There’s a huge number of different models available," confirmed Kelly.
Azure Machine Learning plays into Microsoft’s mobile-first, cloud-first strategy by allowing companies to turn models into cloud-based apps that anyone can use, and you could even work on the model via your Surface tablet, revealed Kelly.
While he declined to share the number of "partners" Microsoft has been trialling the tool with for a year, he did say the firm is committed to "actively building a pretty robust partner network".
He named one as Carnegie Mellon University, using the tool to measure campus buildings’ energy outputs, expecting to save 30% on its bills just by adjusting the air conditioning and heating based on the predictions.
The news comes after a string of big data-oriented announcements from Microsoft in April, including its "big data in a box" Analytics Platform Service.