Unsupervised learning and clusters

Choose your learning level

BeginnerIntermediateAdvanced

Start with plain-language explanations. Switch levels at any time.

Beginner explanation

Unsupervised learning looks for structure without supplied target labels for the main learning task. Clustering groups examples according to a similarity rule. A discovered group is a pattern in the data, not automatically a meaningful real-world category.

Intermediate explanation

K-means partitions points around k centroids by minimizing within-cluster squared Euclidean distances. Feature scaling affects the result. Other clustering methods use different assumptions, so the same dataset can produce different reasonable groupings.

Advanced explanation

Select representations and distance metrics deliberately. K-means favors compact clusters under its objective and can be sensitive to initialization and outliers. Internal scores such as silhouette do not prove business usefulness or ethical validity.

PUT IT IN CONTEXT

A worked example

Grouping fictional customers by visit count and spend may put high-spend visitors together. If dollars range into thousands and visits into tens, raw distances can be dominated by spending unless scaling is considered.

TRY IT YOURSELF

Practice the idea

Points 1, 2, 9 and 10 form two obvious one-dimensional groups. Calculate their centroids and state what the clusters do not tell you.

Your working notes (not saved or sent)

Reveal the answer guide

Centroids are 1.5 and 9.5. These groups do not establish why the points differ, whether the pattern will persist or what action should follow. Interpretation requires additional evidence.

Go further: advanced challenge

Compare cluster stability under different initializations and resampled data. Explain why stability still does not establish causation.

An open-ended discussion task. Explain your assumptions and evidence; there may be more than one defensible answer.

Continue the conversation

What part of this idea would you like to test or question? Compare your reasoning with the community.

Find related forum discussions ↗

Opens a topic search in the existing AstraCruz forum. Results depend on available posts; signing in may be required. No discussion or activity is fabricated.

Browse all forums