Unsupervised learning analyzes unlabeled data without feedback, seeking hidden structure rather than applying predefined categories. Its central task is grouping similar points while separating dissimilar ones. section 12.1.1
Clustering turns the general objective into practical data-analysis tasks:
- Segment customers by purchasing behavior, then inspect the resulting groups to interpret their composition. section 12.1.1
- Find unusual data, such as signal noise or other points that stand apart from the main structure. section 12.1.1 section 12.6
- Preprocess large datasets by breaking them into coherent groups before applying more sophisticated algorithms. section 11.4.4
Customer grouping
Given people described by age, education, financial situation, and similar attributes, clustering can produce meaningful groups. Analysts must interpret those groups; the algorithm does not name them or explain what they represent. section 11.4.4
The choice of clustering method determines which structures are visible. K-means is simple but limited to convex clusters; DBSCAN handles arbitrary shapes but struggles with heterogeneous density; OPTICS supports varying density through hierarchical clustering. section 12.6



















