Microsoft is readying a cloud-based machine learning platform that will enable developers to quickly and easily build predictive analytics into applications.
Due to be available as a public preview from July, Azure Machine Learning (Azure ML) will eventually be offered as a fully managed cloud service that aims to help businesses sidestep the difficulties of building data-driven applications to forecast future outcomes and make better decisions.
Eron Kelly, general manager for Microsoft's Data Platform, told V3 Azure ML will be a key component in the firm's overall data strategy. "We're very excited about this, because machine learning enables you to shift from analysing past behaviour and start to anticipate the future, and that's something we've seen lots of customers really focused on moving forward," he said.
Azure ML brings together new analytics tools with algorithms developed for products such as the Xbox and Bing services, and years of machine learning experience, according to Microsoft.
In this respect Azure ML is similar to IBM's Watson, but Microsoft said it is aiming to bring its machine learning capabilities to a wider audience, including smaller businesses.
"Watson is an IBM Global Services project, it's about ‘open up your cheque book' because it's going to be a pretty expensive arrangement, whereas this is a cloud-based service where with a credit card you can fire up the service and start to get the power of machine learning very easily," Kelly said.
Ease of use is another area that Microsoft is trying to address, so even organisations without resident data experts can derive value from Azure ML. To this end, Microsoft supplies a large palette of pre-defined algorithms for analysing datasets, which users can drag and drop to select and combine together to develop their model.
Customers can also write custom algorithms using the R programming language, while Microsoft is also preparing an SDK that will allow partners to build solutions for specific applications, Kelly said.
One component, dubbed the Machine Learning Studio, enables customers to upload data and train their model to make better predictions. Kelly cited the example of Microsoft's online Store, where you can feed the model historical data of known fraudulent and non-fraudulent transactions in order to train the model to recognise them better.
Once trained, users can use application programming interface (API) calls to feed data to their model hosted in Azure for analysing.
Kelly said that because of big data, customers need access to services such as Azure ML to help them process and analyse information effectively.
"The challenges customers are facing are that there is a huge shortfall of data scientists. Because of this lack of data science skills, having tools that are easier to use and more intuitive is really important, and that was a key design point for us," he said.
Kelly declined to say when the service is likely to hit full production status, but said he was confident it would be relatively soon. The firm has been working with about 100 customers in a private preview for about a year, including some internal Microsoft teams.
Using photocatalysts to convert carbon dioxide into usable energy such as methane or ethane
Trained on curated data from Moorfields Eye Hospital, the neural network also shows clinicians how it reached its judgement
Yokohama National University demonstrate technology that could lead to a fault-tolerant universal quantum computer
Top-of-the-range Threadripper 2990WX now available from Scan, Ebuyer, Overclockers, Novatech and Amazon