K-Means Clustering

Identify natural groups in your data.

Definition

K-means clustering partitions data into k homogeneous groups by minimizing within-group variance. It is the most widely used clustering algorithm in unsupervised learning.

When to use it

• Segment a population into distinct profiles
• Identify natural subgroups in data
• Explore the structure of an unlabeled dataset

Requirements

• Continuous variables
• Standardized data recommended
• Number of clusters k chosen by the user (selection aid provided)

What StatsLab computes

• Assignment of each observation to a cluster
• Cluster centroids
• Within-cluster sum of squares (Within-SS)
• Elbow plot to choose k
• 2D visualization of clusters (PCA)
• Descriptive statistics per cluster

Worked example

Context : Segmenting 200 customers into groups based on purchasing behavior (frequency, amount, recency).

Result : 3 clusters identified: Loyal (n=68), Occasional (n=89), Inactive (n=43)

Interpretation : The elbow plot suggests k=3. Loyal customers spend 3× more than inactive ones. This segmentation guides differentiated marketing strategy.

Run this analysis