How machines learn
Clustering is an unsupervised learning method that groups similar data points together automatically, without being told in advance what the groups are.
Drop a bag of mixed buttons on a table and you will naturally push similar ones together: big ones here, red ones there. Clustering is a computer doing this with data. It measures how alike items are and gathers the alike ones into groups called clusters.
No labels are involved. You may tell the method how many clusters to look for, or let it decide. The results depend heavily on how you measure 'similar,' which is why the same customer data can produce different groupings when you change which features you feed in.
Clustering is used to segment customers, organize documents by topic, compress images, and flag outliers that fit no group. It also helps explore a new dataset before anyone has decided what questions to ask. The clusters still need a human to name and interpret them.
An online store clusters shoppers by what they browse and buy, then discovers a group of late-night bargain hunters it never knew existed.