Unsupervised Learning
Finding structure or groupings in unlabeled data, with no target values to predict.
- Gets only inputs, no labels.
- Outputs: clusters, low-dim summaries (PCA), or anomalies.
- Useful when labels are missing or the goal is exploration.
When to use: When labels are expensive or unavailable, or when the goal is to discover structure rather than predict.
Example: K-means on customer purchase vectors discovers 5 segments (price-sensitive, premium loyalists, weekend shoppers) with nobody labeling examples first.