unsupervised
K-means clustering
Explore whether observations form compact groups across several numeric features.
Requirements
- two or more numeric features
- a defensible cluster count
- distance-based grouping is meaningful
Avoid when
- clusters are elongated or very unequal in size
- categorical features dominate
Formula
arg min Σₖ Σₓ∈Cₖ ||x−μₖ||²
Interpretation
Cluster profiles describe group centers; silhouette measures separation from -1 to 1.
Worked example
Explore compact patient profiles from standardized continuous measurements.
Open example datasets