Data science and artificial intelligence may help service providers speed up service provisioning, provide better network performance and more comprehensive security. But none of that matters if the models used in research labs don't work in the real world, according to researcher Nick Feamster.
Feamster is just starting his new gig as director of the Center for Data and Computing (CDAC) at the University of Chicago. He was previously a professor in the Computer Science Department at Princeton University. One of his jobs now is to find out what kinds of machine learning and data science can be applied in real networks today, at scale.
On this episode, Feamster discusses the everyday tasks service providers can use with machine learning to help broadband subscribers solve their technical issues and avoid calling customer service. With hundreds of encrypted video streams traversing their networks, carriers can't be on the hook for how each application performs, but they can build a model that generalizes across several different services and helps troubleshoot problems before they become customer complaints and a drag on the business.
He also hints at some advanced services -- like how activity recognition could be tied back to a health or security monitoring service. Bridging the gap between having the information and using it is the challenge, Feamster said. As you'll hear on the podcast, measuring networks and gathering data are the first steps, but finding out what data is most important -- and telling the service provider or the customer to take action -- is the key to unlocking new services.
The Light Reading podcast is available on:
Please subscribe today. It's free and worth it.