Have you ever had a credit card transaction declined when it shouldn’t have been? Or been on the receiving end of a personalized email or web ad? Have you ever noticed a site giving you recommendations for things you might be interested in when you're shopping online?

And my last example, have you ever had an offer from a company designed to stop you from leaving them as a customer? If any of these things have happened to you, then you’ve probably been on the receiving end of a machine learning algorithm, employed by a company you do business with (or in some cases, have merely considered doing business with).

An Overview of Machine Learning

We’re going to take you behind the scenes and give you a layman’s view of machine learning so you can see what kind of problems they can solve. If you’re a data scientist, then you might be more interested in this big data journey about accelerating data science, which is more detailed. But this article is designed for technical people who hear the buzzword, who know that it's something important, but don't really know what it is or what it can do.

You'll get just enough information to make you dangerous.

What Is Machine Learning?

A McKinsey article describes machine learning as "…based on algorithms that can learn from data without relying on rules-based programming". Put another way, with big data you've got a lot of data. Determining what do to with it and figuring out what it’s telling you isn’t easy. So you can understand the appeal of machine learning, which basically allows you to find processing power and the right algorithm and tell them to figure things out for you. The analogy is how we learn as human beings, experiencing the world around us and working things out for ourselves.

When I taught my kids how to ride a bike, I didn't give them "The Rules of Bike Riding". I put them on a bike, held onto it/them, and let them work it out. They took data inputs from their eyes, their ears and, on one occasion a large bush, and started to discover what would keep the bike upright. So it is with machine learning. Take the data, work with it and see what comes out.

Uses of Machine Learning

Machine Learning Uses

Supposed you've been tasked with finding out more about your customer base. In the snapshot above, you can see they're a diverse and pretty happy lot. But what else do you know?

Well, a simple query in your database might reveal things like age, gender or how they like to be contacted (mail, email, phone, text).

You could run a query with some analytics to calculate, say, RFM, a measure of customer value based on how much customers spend and how often they spend. You can see who is more valuable to you, but you wouldn't really know what to do with that data.

Machine learning algorithms could do much more. For example, you could group your customers into segments that show similar behavior, or you could also figure out how likely they are to purchase a given new product of yours. In the picture above, we have customers in five different behavioral segments: "retired cosmopolitan", "affluent executive", "new home mom", "young, successful startup" and "executive product collector". And with your data, you might even know their likelihood of purchasing a product that you're promoting. 

Now you have something potentially more powerful. Armed with this information, you can:

  • Tailor your marketing campaigns
  • Use different language for those different groups
  • Prioritize campaigns
  • Market that product only to the subset of customers likely to buy it

Machine learning gives you much more insight into your customers and, perhaps most importantly, it can predict what they might do or respond to. These days, we’re starting to take deep learning to another level and use it to solve real business problems—which is very new and exciting.

But how does machine learning actually do that? Join us again next week, when we take you through some of the most important machine learning techniques you should know about.

Source: Oracle Big Data Blog posts