Machine Learning is the buzzword of the moment. In recent years, news stories raving about its possibilities have soared, Google searches for the term have quadrupled, and companies across the globe have been scrambling to figure out how to capitalize on the excitement by bringing it into their product mix.
While that can be a great thing, claims made by some businesses about what Machine Learning can do are wildly exaggerated. That makes it crucial to cut through the noise and get to grips with its potential, limitations, and what you can realistically achieve with your resources so that any investment makes solid business sense — so say Philip Lima, CEO of Mashey, and Boaz Farkash, Head of Product Management at Sisense. The pair joined forces to deliver an in-depth webinar on Machine Learning and business intelligence, which you can view in full here.
What Is Machine Learning?
The definition of Machine Learning is actually very simple, says Philip. It’s a system that trains itself to come up with the correct output based on the inputs it’s been given.
When you apply for a credit card, you give details like your name, address, and so on, which the Machine Learning application merges with other data such as your credit score. Based on these inputs, the algorithm assigns you a profile, assessing your likelihood of repaying this credit, and approves or denies your application accordingly.
Simple uses of Machine Learning permeate our day-to-day lives. Consider spam filters, which essentially guess whether a message is junk based on how closely it resembles emails that previously earned this tag.
More recently, though, these basic applications have evolved into “Deep Learning,” allowing software to perform increasingly sophisticated tasks with considerable implications for the way we do business.
Everyday Use of Machine Learning
Today, as Philip points out, you can deposit a check with your phone simply by taking a picture of the front and back. The algorithm identifies all the important bits, figuring out the amount, name and account number, verifies that it’s real and unused, and then proceeds with making the deposit.
Or take the phenomenon of your iPhone warning you that it’s time to leave for your appointment, based on how long it thinks this will take under current conditions. For that to work, an elaborate process has to happen, taking in information from your calendar, figuring out your location and likely routes, calculating how much traffic there is, and then, combining all these inputs, output advice on when you should head off.
These are both examples of Machine Learning right under our noses. According to Philip, in 10 years’ time, you’ll be hard pressed to find any tech that doesn’t incorporate Machine Learning – and one of the most intriguing areas where this is the case is — you got it — business intelligence.
The key question to ask yourself is: how do I apply this to my business to make better decisions? Or, put another way: when does it make sense to invest in Machine Learning projects for my business?
Natural Language Processing
One of the most exciting applications, says Boaz, is Natural Language Processing (NLP).
NLP is when software can take natural speech, delivered by voice or in writing, comprehend it, and answer you in much the same way as a human does.
For example, Sisense Everywhere uses bots and NLP to deliver data insights outside of the usual dashboard environment. You can be having a conversation over, say, Skype messenger with a colleague, address a question or request a specific set of stats/graphs/dashboards from the bot, and they’ll contribute this to your conversation seamlessly. You can even ask the bot for a more detailed analysis, and they’ll automatically post a written breakdown of the graphics you’re looking at accordingly.
Think of this like Siri or Alexa, but with high-powered data analysis built in. Instead of asking Google Home or Amazon Echo what the weather’s like today, says Boaz, imagine if you could ask, “Alexa, how are my sales figures this week?” and get meaningful insights drawn from your business’ entire pool of data, just like that.
This is just one of many possible uses of Machine Learning to boost BI, as Philip and Boaz explain in depth in the webinar.
You can, for example, use this to detect anomalies in your BI workflow, getting automatic notifications of blips in your most business-critical KPIs, so that you never miss out on important incidents.
You can also use it to figure out when you need to scale up operations to satisfy demand, basing this on market influences and historical data that indicate how long your lead times are and when there is a surge of demand for your product so that you never miss out on profitable opportunities.
The key is simply to make sure that your chosen application has realistic potential for ROI.
As Philip explains, let’s imagine your company suffers an average of two major tech failures a year, costing around $50,000 each time. If introducing a failure prediction project to your BI activities — one that’s based on Machine Learning — costs you $85,000, even if you only succeed in halving your failure rate, by year two you will have likely saved yourself $100,000, more than making back the cost of your investment.
One to Keep Your Eye On
Whereas most analytics platforms strain under the weight of more data, Machine Learning thrives on it. The more inputs you can pour in, the more evidence the algorithm has to adapt and refine its understanding. In other words, the faster it can “learn.”
This makes Machine Learning uniquely suited to a BI landscape. While the applications aren’t yet perfect, the tools are getting smarter and smarter all the time. We haven’t even scratched the surface of the potential of this powerful technology — and when we do, we guarantee you’ll want to be ready and positioned to take full advantage of what it can do.