Supervised learning trains a model on labeled examples with known correct answers, while unsupervised learning finds patterns in data that has no labels at all.
The core difference is whether the training data comes with "answers" attached.
The model learns from input-output pairs — for example, thousands of emails each labeled "spam" or "not spam." Once trained, it can predict the label for new, unseen emails. Most practical business applications — fraud detection, price prediction, image classification — use supervised learning.
The model only gets input data with no labels, and must find structure on its own — like grouping customers into segments based on purchasing behavior, without being told in advance what the segments should be (called clustering), or discovering which product purchases tend to occur together.
A third category, reinforcement learning, sits apart from both — an agent learns by trial and error, receiving rewards or penalties for its actions.
Last reviewed: September 2026