Ever wondered how your iPhone knows where you live, or how long it will take you to get from home to the office?
Machine learning is so ubiquitous you have probably used it several times today without even knowing. This phenomenon is transforming the way we interact, and depend, on technology, and it’s the hottest tech trend today.
Machine learning helps computers or devices to understand, learn, predict and even operate autonomously. It does this by looking at data and developing a statistical model, or algorithm, that can predict patterns of behaviour. At its most advanced, it uses voice and image recognition to predict your next move. This is known as deep learning, a form of machine learning. Here’s a good example: have you uploaded a photo only to have Facebook suggest whom you should tag in the image? This is because Facebook has an inbuilt facial recognition algorithm that predicts your friends faces based on facial features, like the distance between the eyes, nose and ears. Clever stuff.
While many companies are embracing machine learning to improve products and services for their customers, many are still trying to understand how exactly to use it. The good news is, once a competitive advantage for companies armed with the brightest data scientists, there are plenty of open source tools available. Programs like AWS’ Amazon Machine Learning and Google’s Mainstream Machine Learning make it easy for developers of all skill levels to use machine learning technology. So as long as you have a computer and Internet connection, it’s possible to get started. The most important thing is getting started. Before long machine learning will move from realm of a brave few in the tech team to something everyone is doing.
Regardless of whether you choose to enlist your own data scientist or work with existing programs like Google and AWS, there are four key phases in machine learning you need to know about:
Data, data and more data
Data underpins the digital economy. The Global Review of Data-Driven Marketing and Advertising revealed more than 80 percent of marketers (surveyed from 17 different global markets) think data is pivotal to their advertising and marketing efforts. The more data, the better – machine learning works when there is an abundance of data to leverage. Things like: the physical addresses of customers, their location when they logon to your website, which pages they exit, what time of day they buy your product, whether they are married or single, and which website they were referred to your site from.
Someone to make sense of the data – like a data scientist
Generally teeming with creativity and an aptitude for numbers, data scientists make discoveries while swimming in pools of data. While looking at logs of data, they recognise how users interact with a product or service to better understand their needs and desires. These are usually the people helping Netflix recommend shows you might like, or programming Amazon to suggest books you should buy. And they’re hot property right now; Harvard Business Review ranked data scientists as the sexiest job of the 21st century.
Then comes the algorithm
Once patterns in customer behaviour have been identified, data scientists try to find ways to automate that process.
Using existing data, data scientists develop a model that maps inputs to outputs. For example, a retailer may look for a correlation between the temperature and jumper sales. An algorithm is designed to test this pattern.
Making sure the algorithm works
This algorithm linking weather with jumper sales is repeated, over and over, until it is as accurate as possible. Then, rather than tracking what has happened (i.e. the temperature drops to below 15 degrees) and responding in real time (order more jumpers), machine learning goes one step further and allows organisations to make better predictions about what will happen (i.e. the temperature will drop to 15 degrees next month, time to start marketing to customers who might buy a jumper).
Here we have a Machine Learning model
A good data scientist can optimise results so that an algorithm can predict the right outcome even when using new data sets (new customers, new locations). This is where the weather example becomes obsolete – not even the best meteorologists can seem to predict the forecast, but you understand the analogy I’m sure.
There are plenty of other great examples of machine learning and the benefits for business. Leading health management company, Medecision, developed an algorithm that was able to identify eight variables to predict avoidable hospitalisations in diabetes patients. Surprisingly, the analysis found that most of the avoidable hospital admissions depended on whether the patient had a flu vaccine. Most hospitalisations were due to upper respiratory infections that were complicated by diabetes, but not caused by it.
Use data wisely
By combining the right data with talented data scientists, the opportunities for business are endless. However, this comes with a caveat. As with any involvement with personal information, it is incredibly important to exercise responsibility when using personal data to benefit your business. Last November’s Senate inquiry into insurance found companies were increasing premiums based on the time of day, the colour of a vehicle, or marital status of a customer. While this may have increased revenues, it could also be considered rather unethical.
It’s scary to think less than five years ago, Google had just two deep learning projects. Now set in motion, it is moving at lightening speed. In the next five years, artificial intelligence will transform society. It will change what we do, versus what we ask a computer to do for us. Most certainly, the best is yet to come.