If you're just starting to get into data science and machine learning, you've probably heard of unsupervised learning a lot. It is generally used for classification problems. When classifying sets of data, you have a question that you need answered. Say for example you only have today's temperature, humidity, air pressure, and inches of rain today. And you want to try and predict if it will rain tomorrow. When you look up historical data, your targets become the answer to whether or not it rained the next day. If yes, then your target is a one. If no, then your target is a zero.
Unsupervised learning is when you don't set targets for your machine learning algorithms to learn from. And in this example, it would be if you didn't know if it rained the next day. The advantage of unsupervised learning is that it might be able to pick up on events that don't normally happen, like a hurricane.
Another example of this is in fraud detection. Fraud detection algorithms are pretty good at detecting common frauds. However, frauds make the most money by inventing new methods, which regular supervised learning cannot pickup.
Unsupervised learning has a lot of potential ranging anywhere from fraud detection to stock trading.
Top comments (0)