k-means clustering

K-means Clustering

An unsupervised ML algorithm that groups data points into k clusters by minimizing the distance between each point and its cluster's center — used to find structure in unlabeled data.

When you'd see it: Customer segmentation (grouping customers by behavior patterns without pre-defined categories), market research (identifying distinct audience segments), anomaly detection (points far from any cluster center are outliers), and initial data exploration.

Why it matters: K-means provides a data-driven segmentation without the researcher imposing a pre-determined scheme. It can reveal natural groupings in customer behavior, feature usage, or geographic patterns that intuition would miss.

Common mistakes: Choosing k without justification. The 'right' k isn't in the data — you have to choose it. Use the elbow method (plot inertia vs k and look for the bend) or silhouette scores to guide the choice. Also: k-means assumes spherical clusters of similar size — poorly suited for irregular or very different-sized clusters.

Related principal component analysis, normal distribution, outlier