Predictive analytics is a familiar concept, though not so much when it comes to network management, performance optimization, and capacity planning. Thanks to advancements in artificial intelligence and machine learning, predictive analytics has evolved into an invaluable tool for enterprise networks. We talked with John Crupi, vice president of IoT analytics for Greenwave Systems, about the latest developments in predictive analytics.
Where are we with predictive analytics at this point in time?
Artificial intelligence is having a big impact. It is a matter of what we can learn from the data, what patterns AI can extract out of this data to give us insight into how things are performing. Visual analytics and pattern detection allow us to look at behavior and understand the patterns, then use that understanding to drive the things that represent network performance and anomaly detection, for instance. We can make it autonomous through machine learning, with the human side driving what is of interest.
What has changed in recent years to make predictive analytics so useful now?
It is the wealth of open source tools and the ease of using the cloud to process an immense amount of data. This has been driven by IoT, where you are talking about billions of devices being connected. There is a real-time connectivity to those, and you are looking at perhaps millions of network devices. You need systems that can analyze the data for you. The notion of understanding the past to predict the future involves a complex set of pattern detection, and the best way to do that is with systems that continually learn about behavior in order to predict, and refine, where things are going. It’s not always just about detecting problems, but knowing when they are starting to happen and hopefully have them self-heal.
How extensive is the current use of predictive analytics by enterprises?
It depends on the industry. If you look at internet and ecommerce companies, analytics are core to their operation because they have such a tremendous amount of data. But for other companies, it is still early. Before you can leverage predictive analytics you need insight, and many companies don’t have that. We have talked to companies with million-dollar surgical robots, yet those robots aren’t even fully connected. So while there are a small number of companies embracing predictive analytics in the industrial space, we see that growing exponentially soon.
Do you find CIOs generally accepting of the predictive analytics concept?
At this point, a large proportion of CIOs who are forward thinking see the value and know this is real. The struggle for them is that it is not their domain expertise. The sticking point is where to begin and how to get the help they need.
How do they get started?
It requires data scientists, working with domain experts in what you are going to analyze. The domain experts are the network engineers and the people in network operations. They know when there are issues and problems, and they want faster insight into what they need to understand. Data scientists are the facilitators, and create the models, but they need to know the important elements in the network, which is what the network engineer or analyst determines.
What specific steps should an enterprise take, in light of the fact that there is more demand than supply when it comes to data scientists?
You create a tiger team that includes a data scientist, and if you can’t find one to hire, there are companies and consultants that can help you. Start with a small project, stay focused, and continually iterate and understand the value you are getting from it. Determine the most important data elements, what the problems are, and what kind of future insight you want. Where CIOs struggle is that this may sound “cool” but they’re not sure how much effort to put into it. As you progress, build a use case and proof of concept. Consider whether some of the things that predictive analytics makes possible are just nice to know, or if they can be transformative to your business.
Is this an area where machines will crowd out the human element?
No, and the reason I say that is when you look at the world of robotics and the military, for instance, there will always be a need for a trigger puller. Machine learning and artificial intelligence are good at processing tons of data, taking out the mundane work. But it is so humans can see patterns that were not otherwise visible. I don’t see this displacing people, but instead giving us greater ability to build efficient and complex networks.
Anything else CIOs should know?
It is time that they really start taking machine learning, artificial intelligence, and predictive analytics seriously as applied technology, as a part of normal operations. I think that in the future CIOs will have analytics teams, building systems that are constantly learning, analyzing performance, and building expertise into the infrastructure and the company.Back to all blog posts