Unisys today launched its machine learning–as–a–service solution, designed to enable clients to predict changes in business conditions and determine appropriate responses to those changes. The new service, which is a component of the Unisys Analytics Platform, uses machine learning to power predictive and prescriptive analytics—specifically, pairing a library of machine learning algorithms with a suite of processes for analyzing clients' data.
There are four analytics stages, according to Rod Fontecilla, vice president and global lead for analytics at Unisys. "The first stage is to be able to ingest this massive amount of data, making sure that you have the infrastructure and the capabilities to ingest it," Fontecilla says. "The second stage is to start analyzing the data…doing what nowadays people are calling data wrangling, making sure that you understand what the data is, where the data is coming from, and how you can correlate external data with your internal data."
The third and fourth stages are where predictive and prescriptive analytics come in. "When you go into the predictive stage...you build predictive models using machine learning algorithms to make better business decisions today that are going to affect the future. Once you get to that stage of predictive modeling, you are at a point where you can really make a big impact in the business. You get into the prescriptive side where, due to these predictive models and understanding what is going to happen in the future, you can optimize your business processes; you can start automating your workflows."
Fontecilla also emphasized the ability of machine learning to dynamically improve business rules. "The massive amount of data and the computing power that we have at our fingertips have made the advances of [machine learning] algorithms very significant nowadays," he says. "Until now, for the most part when you’re building applications or you’re running statistical models or you’re building rules engines and things like that, you [would] hard-code the rules—the rules get defined by the business, they get hard-coded, and usually you revisit the rules once a year."
But with machine learning, the technology become responsible for the rules. "The beauty of it is that these algorithms get smarter on their own with the more data you throw at them, that's the reason for the name 'learning,'" Fontecilla says. "And the rules, you don't need to define them anymore or hard-code them anymore; the rules are automatically created by the algorithm itself.”