Clustering

Clustering

Discovering natural groupings in data without predefined labels.

Learning Outcomes

By completing this topic, you will:

• Apply K-means, DBSCAN, and hierarchical clustering
• Choose the optimal number of clusters
• Interpret and validate cluster results
• Create customer personas from segments

Visual Guides

K-Means Algorithm

Elbow Method

Silhouette Scores

Prerequisites

• Unsupervised Learning concepts
• Distance metrics (Euclidean, Manhattan)
• Feature scaling techniques

Key Concepts

K-Means Clustering

Partition data into K spherical clusters:

1. Initialize K centroids
2. Assign points to nearest centroid
3. Update centroids as cluster means
4. Repeat until convergence

Choosing K: Elbow method, silhouette score

DBSCAN

Density-based clustering for arbitrary shapes:

• Epsilon: Neighborhood radius
• MinPts: Minimum points to form a cluster
• Automatically detects noise points

Hierarchical Clustering

Build nested cluster hierarchy:

• Agglomerative (bottom-up)
• Divisive (top-down)
• Dendrograms for visualization

When to Use

Common Pitfalls

• Not scaling features before clustering
• Choosing K based on convenience, not data
• Ignoring cluster interpretability
• Using K-means on non-spherical data
• Forgetting to validate cluster stability

(c) Joerg Osterrieder 2025