3 Machine Learning Technologies to Watch Over the Next 3 Years

Kirk Borne

Kirk Borne

What's new and what's on the horizon for machine learning and analytics? We asked Kirk Borne, Principal Data Scientist and Executive Advisor at Booz Allen Hamilton, what machine learning technologies he's watching. He focused his reply on applications, not algorithms.

"When I think about what's new and coming up, I don't expect that it's the mathematics that will be changing but the types of things we can do with the math," he says. With that in mind, consider these three applications of analytics that Borne is watching.

1. Graph analysis finds connections

Graph analysis is what we used to refer to as social network analysis, but then Facebook and Twitter came along and the term "social networks" took on a whole new meaning. Here's a broad explanation from Borne.

"Just about everything in the world has some relationship to something else," he says. "You have a family, and you work at a business. That business has employees, and it has customers. Looking at your connections alone, you can tell there's all kinds of connectivity going on in the world."

Graph analytics is the process of discovering patterns in these networks. Some of these patterns can be helpful in running a business or improving a process, and some patterns can detect crimes or other preventable behaviors.

"Crimes like money laundering and insurance fraud are all perpetrated by networks of people who are laundering money or submitting insurance claims that might look legitimate on the surface, but aren't," explains Borne. "Understanding connections that are maybe two steps or three steps down the chain can help identify the bad actor or the person who is perpetrating the fraud."

Graph analytics has positive and negatives applications. "Negative applications are the things we're trying to prevent, and the positive applications are the things we're trying to produce, like improved healthcare, improved education, improved transportation or improved energy use," says Borne. "Understanding how things connect and relate to one another is all built around that sort of graph model, where you discover how A is linked to B, and B to C, and C to D in our world."

Learn more about fighting fraud with graph analysis (also known as network analysis).

2. Geospatial analysis adds new dimension to analytics

The second application of machine learning to watch in the future, according to Borne, is geospatial analysis.

"Everything we do takes place in time and space. And so the tools we're building, in terms of visualization and prediction, ought to take the contextual information of space and time into account," says Borne.

For example, do patterns of behavior change based upon time and location? Are you more likely to purchase items on Fridays? And does your location influence your buying habits, your health decisions, your energy usage, your interests, your entertainment choices, your online preferences and so on?

"Building models that have that sort of spatial and temporal dimension to them is becoming more common. When you add those aspects of data and algorithms, it's dealing with really complex data."

For more on this topic, read, Location Analytics: Why Adding "Where" Makes BI Better.

3. Natural language generation can create and tell your data's story

Natural language generation has become commonplace through virtual assistants on our phones and in our homes that can interpret our voice commands and reply in the same language.

Taking that capability a step further, Borne explains how NLG techniques can be used to help generate reports.

"In many cases, the report does not yet exist but the database exists. Natural language generation can automatically generate a written report for you from the database. Based on your voice prompt, the system finds the patterns, trends, anomalies, and gaps and generates a brand new report from the data," explains Borne.

As data volumes continue to grow, and personal assistants become the norm, technologies that can help us interact naturally with data will become even more popular.

"If we can have a chat bot tell us the latest news, the stock market and all the financial numbers, the next step is for it to tell you the story behind the numbers in an interpretable way," says Borne. "I think that's really valuable."

Learn more about natural language generation in the article,Your Personal Data Scientist.

What machine learning technologies are you watching, and how do you see machine learning changing the future? Tell us in the comments.

This content was reposted from SAS Voices. Go there to read the original post.

Alison Bolen, Editor of Blogs and Social Content

Alison Bolen is an editor at SAS, where she writes and edits blog content and publishes the Intelligence Quarterly magazine. She recently picked up and moved her family and her home office from Ohio to New York. Since starting at SAS in 1999, Alison has edited print publications, Web sites, e-newsletters, customer success stories and blogs. She has a bachelor's degree in magazine journalism from Ohio University and a master's degree in technical writing from North Carolina State University.

How Can Marketing and IT Work Better Together?

If your marketing team is running up against obstacles when dealing with IT, here's a roadmap for communications to help get projects moving again.

3 Machine Learning Technologies to Watch Over the Next 3 Years

Booz Allen Hamilton's Principal Data Scientist and Executive Advisor Kirk Borne recently weighed in on a few of the rising technologies for machine learning. Here's what he said.

Re: Natural Language Reporting
  • 9/30/2017 7:11:14 PM

Carry on. Nothing to see here. Nobody's watching...worry not! ;)

Re: Natural Language Reporting
  • 9/19/2017 7:57:41 AM

Kq4 writes that "It will be interesting to watch machine learning to see if it may help us to see more of what's really happening in patterns and trends instead of our sometimes false human hypotheses."

I agree – it would really be nifty to have an AI/ML app that could review accumulated research data (or even accumulate the data for you), then crank out a complete report with insightful analysis for you. I'd imagine this would be particularly nifty for highschool and college students.

The other two emerging ML technologies also sound interesting. They would detect pattern linkages within all my activities, and keep tabs on where and when I am. This would be useful to ... hey, waitaminit ... 


Re: Natural Language Reporting
  • 9/10/2017 9:01:49 AM

I am sure there are plenty of fales hypotheses. We may verify some as well.

Re: Natural Language Reporting
  • 9/10/2017 8:03:03 AM

Humans like to find patterns and that can pose some problems where we try to attribute ceratin causes to a rssult we see. It will be interesting to watch machine learning to see if it may help us to see more of what's really happening in patterns and trends instead of our sometimes false human hypotheses.

Re: Natural Language Reporting
  • 9/8/2017 8:25:42 AM

I think adding schemas to the natural languauge would provide another level of usefullness. The ability to make asscoiations and connections really ups the game.

Re: Natural Language Reporting
  • 9/7/2017 9:12:16 AM

For specific industries or as you mentioned franchises that does sound feasible.  That model should work well even without the natural language system.  I think it's more likely that the big DB vendors will start working with companies to build out the relationships needed to make the natural language queries work in their environments.  I see a niche of analysts popping up just to handle this kind of work to dissect data, build relationships and translate them to natural language based on questions most likely to be asked.


Re: Natural Language Reporting
  • 9/6/2017 1:32:51 PM

A large franchiser could make a natural language system available to it's franchisees who operate one or several stores.

With standard programming from corporate, an initial vocabulary could be created. Each store owner or manager could enquire about various topics. How are sales or profits trending? Which items should be re-ordered?

With specific store data loaded in each machine, the store would have good and timely information. Meanwhile, the full network of machines could improve based on what new questions were being asked in multiple locations.

Natural Language Reporting
  • 9/6/2017 8:45:34 AM

I'm really interested in seeing how natural language generation is implemented.  I imagine the frustration of someone trying to ask natural language questions for fields that don't exist or are calculated from other fields.  A large part of writing a story with data is knowing how each chunk of data relates to another.  Someone has to define those or know how they are related before you can even begin to look at asking natural language questions.  Once those definitions are built you'll still have a lot of natural language questions that need a great deal of interpretation. For pre-built questions I can see this being helpful but I don't see it doing away with query languages any time soon.