The most widely used methods of machine learning are supervised learning, unsupervised learning, and reinforcement learning. In this case, “supervision” simply means that the answers are known during the training phase of machine learning model development. Supervised learning is commonly used in applications where historical data predicts future events. A good example of this is fraud detection. The system can detect and flag a fraudulent transaction, and it can be verified by a user who knows whether or not it was indeed fraud. Ever have your credit card company report possible fraud only to realize that it wasn’t (or vice versa)? This goes back to what was mentioned above—the computer is not always right and can produce false negatives and positives. Other examples of supervised learning include image classification, weather forecasting, life expectancy estimation, and diagnostics.
Unsupervised learning, as you might guess, is used against data with no target or label values, so the system does not have the “right” answers. The goal is to explore the data and glean meaningful insights within. Examples of unsupervised learning include targeted marketing, customer segmentation, structure discovery, and big data visualization.
Finally, reinforcement learning, is a type of machine learning that uses trial and error to discover what actions maximize rewards. It is often used for robotics, gaming, real-time decisions, and navigation. Self-driving cars? Yep, that’s reinforcement learning at work.