K-Means

← Back to the article

K-Means groups dots by putting them near the closest center point.

What it does

K-Means groups dots by putting them near the closest center point.

Cluster: A group of similar things • Centroid: Middle of a cluster

How it works

  • Choose how many groups you want.
  • Place center points.
  • Move each center to the middle of the dots assigned to it, then repeat.

Uses and applications

  • A store can group shoppers with similar habits.
  • A classroom app can group students by the topics they need to practice.
  • A photo app can group similar colors in an image.

In short

It finds hidden types of items in a big list.

Related methods

  • DBSCAN: Finds groups by density, so it handles irregular shapes and noise.
  • PCA (Data Shrinker): Often used to reduce dimensions before clustering.
  • KNN (Neighbors): Uses distance too, but to label points from known examples.

Back to Patterns, anomalies and recommendations · All topics

The interactive demonstration and the Python walkthrough for this topic need JavaScript.