Clustering
Grouping data points so that points in the same group are more similar to each other than to points in other groups, without any labels.
Also cluster analysis
Sorting a pile of socks into pairs and colors without anyone telling you what colors exist.
Where it fits
+All relationships as a list
Is a kind of / part of
Siblings
Applies to
How it works
Clustering is the most common unsupervised task. Different algorithms define “a group” differently:
- Centroid-based: each cluster has a center; points join the nearest one. k-means is the classic.
- Density-based: clusters are dense regions separated by sparse ones (DBSCAN). Finds odd shapes and labels outliers as noise.
- Hierarchical: repeatedly merge the two closest groups, producing a tree of clusters you can cut at any level.
Because there are no labels, evaluating a clustering is partly a judgment call: do the groups make sense to someone who knows the domain?
Everyday use cases
- Case 01
Shopper types
Group supermarket customers by what and when they buy.
- Case 02
Organizing photos
Group similar faces in a photo library before you name anyone.
Strengths, weaknesses and when to use it
Strengths
- No labels needed.
- Great for exploring a new dataset.
Weaknesses
- The "right" number of clusters is often unclear.
- Results depend heavily on the distance measure and feature scaling.
When to use itWhen you want to discover natural groups in unlabeled data.