Unsupervised Learning
Unsupervised learning finds structure in data without supplied target labels for each example. Tasks include clustering, density estimation and learning compact representations.
A retailer might group customers by purchase frequency and spending, without assigning customer segments beforehand. A clustering algorithm forms groups from the chosen features; analysts inspect and name the groups afterward. Choosing features and a distance measure still shapes what “similar” means.
The resulting groups are not automatically real customer types. An algorithm can split a continuous pattern into neat clusters, or group records by an irrelevant feature. Check whether groups are stable, interpretable and useful for the business question rather than assuming their existence proves a discovery.
Self-supervised learning also avoids manually labeling every example, but creates prediction targets from the input itself. It is often treated as part of the broader unsupervised family. State the objective when comparing methods: discovering customer groups and predicting missing text require different evaluation.
Sources
- scikit-learn: Clustering — Explains clustering unlabeled data, input representations, method assumptions and evaluation limitations.
Go deeper
- scikit-learn: Silhouette analysis for KMeans docs
Compare possible cluster counts and inspect how silhouette scores reflect separation.