Image credit: PixabayImage credit: PixabayWonder how Amazon understands you enough to recommend future purchases (pretty accurately) or how Netflix gets you so much so that they know what you might want to binge watch next? Are the folks at Netflix and Amazon snooping through your credit cards or watching you through your window at night as you settle in for some prime Netflix time? Kind of….

Netflix and Amazon are able to “spy” on you thanks to machine learning.

Machine learning is, in its most basic form, a way of teaching a computer to think like a human—by classifying data and making connections and associations between items—but more quickly and accurately than humans and without the preferences and emotions that often color how humans come to conclusions.

How It Works

Machine learning is a function of most, if not all, computers. Often confused with artificial intelligence, machine learning relies on data to either teach the computer something about that data or to recognize patterns and relationships between data all without being programmed to a result.

As a rudimentary example of machine learning, let’s consider a stack of photos (the data). After feeding the data to the computer, a human labels the data for the computer (these are pictures of cats; these are pictures of dogs). After repeated exposure to the images, the machine learning algorithm will begin to recognize the differences between the images of cats and dogs. If we further instruct the algorithm to select only images of dogs, eventually, after repeated exposure, the machine learning algorithm will retrieve only the images of dogs while discarding the images of cats.

Adding in additional dog and cat images will help fine-tune the machine learning algorithm (regardless if correct or not) until achieving accuracy, which comes with the repeated exposure to the images.


While there are different types of machine learning, the most commonly used and recognizable are supervised and unsupervised.

Supervised learning teaches a computer by labeling data. For instance, in the machine learning example of the dog and cat images detailed above, the computer has been “told” what the images are (i.e.: “this image is of a dog”; “this image is of a cat”). Supervised machine learning resembles how you would identify objects to a baby.

Unsupervised learning deals mostly in unlabeled data. Instead of relying on labels to learn something, it is an attempt to recognize patterns without the benefit of labels. Consider the Amazon and Netflix examples previously mentioned when trying to imagine how unsupervised learning works.

Imagine that you spent last night watching a Julia Roberts’ film on Netflix. Based on that data, today you could expect Netflix recommendations heavy in both romantic comedies and Julia Roberts. Unsupervised learning considers the data (your viewing history) and looks for connections, patterns and similarities with that data (other movies).

Real-Life Uses

Examples of machine learning at work can be found in everyday interactions and seemingly simple actions.

Search engines: Consider a simple Google search. Each time you search something in Google, you are witnessing a function of machine learning. A machine learning algorithm produces a list of the top results of your search term in descending order of popularity. Whatever links you click or don’t click will influence how the machine learning algorithm will present the list the next time someone searches that same keyword. For instance, if you skip past the very first link that Google recommends, that link may appear lower on the list the next time someone searches for that term.

Cybercrime and fraud prevention: Machine learning can be used in the fight against cybercrime. Because malware files share similar codes, a machine learning algorithm can be trained to identify similar codes, thereby predicting and preventing Malware files from wreaking havoc. Similarly, machine learning can be employed by credit card companies to detect fraud. A machine learning algorithm can be put to work noticing patterns in spending histories, sending an alert when a transaction doesn’t match the patterns of an account holder’s transaction history.

Transportation: Machine learning is commonly used in transportation—from optimizing bus or train routes, determining traffic patterns and in planning and scheduling. Machine learning ensures the efficient management of transportation systems.

Real estate: Real estate companies often use machine learning algorithms to predict housing market values based on data such as number of rooms, square footage, zip code, inflation, value of comparable homes in the neighborhood and the home’s age.

Autonomous cars/self-driving vehicles: Self-driving cars use machine learning algorithms to personalize automobile settings. Because of machine learning, your car understands what your preferred temperature setting is, the kind of music you like to listen to and how you like your seat adjusted.

Finance: Machine learning can be used to predict anything from stock market trends to a loan applicant’s creditworthiness.

Healthcare: Diagnostics and predictions about a person’s health can be made using machine learning algorithms that make assumptions based on risk factors (i.e.: medical history, geography, etc.).

Social media: Machine learning is hard at work on your social media platforms, whether it is recommending “people you may know” or recommending news stories based on your network activity.

Music: Like your Netflix and Amazon accounts, your Pandora and Spotify accounts are able to recommend music you might like or tailor playlists to your tastes based entirely on a machine learning algorithm's assumptions based on your listening history.


Want a hands-on explanation of machine learning? There are a number of machine learning projects available for beginners. Some of the more popular ones include machine learning programs that allow beginners to bet on sports, talent-scout players or act as a team’s general manager, all using available data about players and teams. There are also machine learning projects that let beginners teach neural networks to read handwriting or to write algorithms from scratch.

Machine learning algorithms also exist that allow you to play the role of detective. Using real-life data, beginners can investigate real-life scandals such as Enron. The algorithm uses data such as the emails leading up to the scandal.

Want to predict a stock’s future performance? Machine learning can help with that, as well, by using data such as prices, volatility indices and history to make predictions.


Aylien—Machine Learning Basics: A Guide for the Perplexed

EliteDataScience—6 Fun Machine Learning Projects for Beginners

LiftIgniter—What Is Machine Learning? The Basics

Monkey Learn—A Gentle Guide to Machine Learning

Wikipedia—Machine Learning

To contact the author of this article, email