This article is more than 1 year old

Your machine used to crunch numbers. Now it can chew over what they mean, too

Moving beyond analytics

Promo Artificial intelligence can solve all your problems. It can raise your children, argue with idiots for you on Facebook, order you a curry from the shop down the road, and even give you a foot massage while you’re eating it.

Well, not yet, but probably soon, if the tech elite are to be believed. In the meantime, you can content yourself with using it to find new insights in your business numbers. We’re told that AI and analytics go together like data clustered ham and regression-analysed cheese. But how does one complement the other exactly, and what extra do you get for your money?

Like most technologies, AI has a big problem: as it gains marketing appeal, so many people begin using the term that it can lose its meaning. Slapping an AI label on something gives it a magical sheen that marketers hope will translate into more sales.

This can happen with analytics if we’re not careful. What’s the difference between an analytics program that simply throws lots of computing power at a bunch of numbers, and an algorithm containing some kind of magic sauce?

“Data mining and analytics deals more with structured data,” maintains Debbie Landers, IBM Canada’s vice president of Cognitive Solutions. “Data mining aggregates this structured data from various sources and starts correlating it to find patterns that humans may not have seen.”

That’s just automated number crunching. Computers have been doing this for years. Before analytics was sexy, people called it business intelligence. No secret AI sauce there, then.

“These AI systems can also work with unstructured data and ambiguity that differentiates it from the pure data mining and analytics,” she explains.

So it’s the unstructured part that the AI applies to. This makes sense, because the branches of AI gaining most traction today – machine learning and deep learning – typically have non-deterministic outputs. They’re "fuzzy", producing confidence scores relating to their inputs and outputs.

This makes AI-based analytics systems good at analysing the kind of data that has sprung up since the early 2000s; particularly social media posts. You’ll see companies applying sentiment analysis to it, where AI algorithms "read" Twitter feeds and try to work out whether ‘LOL bae dat #Beyonce tune be SIIIIICK” is a good or a bad thing.

“Combining analytics with AI can help workers in the finance sector by bringing unstructured and structured information together,” explains Landers. She cites wealth managers as an example. Using conventional structured analytics, a financial advisor might visualise a customer’s trades and investments over time, offering useful information about their investment yields and the level of attention they pay to their account. That might help when rebalancing their assets.

An AI system might go further, though, exploring emails written to the advisor, along with social media posts on the client’s account, and could even factor information about life events that the client is experiencing.

“Now you’re expanding that data set outside of the organization. Then you start layering in things like personality insights,” says Landers. So an AI-enhanced analytics system might surface client sentiments that they aren’t revealing to the advisor, or perhaps even aware of. Sentiment analysis is easy when you’re gawping in disbelief over the latest Trump tweet. It’s more difficult when you’re preparing a thousand client meetings for your army of wealth management professionals that week.

A dynamic segmentation algorithm might use all of this data to understand just how aggressive that client is likely to be in their trades, and suggest ETFs and mutual funds accordingly. “It all helps the wealth manager build a better picture of their customer without spending hours poring over their communication history,” suggests Landers. “You’re trying to understand more layers of what that person’s relationship is with your organization and what it could be.”

The ability to work with unstructured data is one characteristic of machine learning, but we can also apply this branch of AI purely to structured data sets, explains Mark Whitehorn, professor of analytics at Dundee University, who also writes for The Register.

“Data mining is perfectly capable of doing machine learning,” he says, using cluster analysis as an example. Business intelligence tools have used cluster analysis for years, finding clumps of data points close together to identify strong sales of Burberry caps in Romford, say. That isn’t machine learning.

“But say to it ‘here are naughty people who have defrauded insurance companies, and here are their characteristics,” he says. “Build clusters and remember the mathematical description of those clusters. Not where the individual points were, but remember the shape of that.”

Then, show a machine algorithm similar examples of honest customers, and ask it to remember the shape of the data clusters describing those people. “Then you say ‘here are people we don’t know. Which cluster do they fall into?'”

Finding likely insurance risks for further investigation based on learned patterns is a good example of where AI and analytics intersect, with nary a tweet or semantically mined email in sight.

Using machine learning in combination with analytics can also take companies beyond basic insights in other areas such as information security, explains Nick Patience, founder and research vice president of software at analyst firm 451 Research.

Applied to information security, conventional analytics can help IT leaders to find their biggest risk areas and work out where to invest their cybersecurity dollars, but this leaves them tracking a moving target.

“People are constantly buying security products to fix a problem or get a patch to update something after it’s already happened, which you have to do, but that’s table stakes,” he says. Machine learning is good at spotting things as they’re happening (or in the case of predictive analytics, beforehand). Their anomaly detection can surface the ‘unknown unknowns’ - problems that haven’t been seen before, but which could pose a material threat. In short, applying this branch of AI to security analytics could help you understand where attackers are going, rather than where they’ve been.

What does the future hold for analytics, as we get more adept at using them? Solutions are likely to become more predictive, because they’ll be finding patterns in empirical data that people can’t spot. They’ll also become more context-aware, using statistical modelling and neural networks to produce real-time data that correlates with specific situations.

We’re already seeing examples of this with services like travel analytics system Wayblazer. The service scans information from multiple sources including travel reviews and visitor searches, building them into a travel graph that tailors travel results for the firm’s customers.

Naturally, IBM expects that adoption will increase. Landers anticipates that combining machine learning and analytics with natural language-driven interfaces will help drive the uptake of analytics among a group of users that haven’t used them before. IBM prefers the term “Augmented Intelligence” rather than Artificial Intelligence, given the focus of its Cognitive Computing initiative is leveraging these capabilities to extend and enhance (rather than replace) human performance.

“Traditionally many line of business have shied away from math,” she says. “Now, people who had run their businesses based on gut instinct and understanding human behaviour are leveraging things like analytics and Cognitive Computing to enhance their capabilities and improve their competitiveness.”

That’ll be a big step for many people who flee at the sign of an open spreadsheet. If analytics practitioners get it right, though, it could open up new insights for businesses.

More about

More about

More about


Send us news

Other stories you might like