What Is Unsupervised Learning
Unsupervised learning works with data that has no labels. There is no teacher and no answer key. Instead the model searches for hidden structure in the inputs alone.
Two of the most common goals are:
- Clustering groups similar examples together, such as segmenting customers by behavior
- Dimensionality reduction compresses many features into a few that capture most of the variation
Without labels, there is no single correct output to chase. The model relies on a notion of similarity or density to decide what belongs together. Algorithms like k means cluster points around centers, while methods like principal component analysis find the directions of greatest spread.
Because no human annotation is required, unsupervised learning can use the vast amounts of raw data that organizations already collect. That is a major practical advantage over supervised methods.
The tradeoff is evaluation. With no ground truth, judging whether the discovered structure is good often requires human inspection or indirect metrics. The patterns are real, but their usefulness depends on the question you are asking.
Key idea
Unsupervised learning discovers patterns and structure in unlabeled data through similarity rather than a labeled answer key.