Learn to find structure in unlabelled data with scikit-learn — group similar records with k-means, hierarchical clustering and DBSCAN, judge the result with proper metrics, then compress and visualise high-dimensional data with PCA and t-SNE/UMAP. You will finish able to take a raw feature table and honestly answer "what natural groups and low-dimensional structure live in here?".
Watch the free preview
What unsupervised learning is, and when it helps — free to watch, no account needed.
What you'll learn
- Find structure in unlabelled data using clustering and dimensionality reduction in scikit-learn
- Fit and tune k-means, hierarchical clustering and DBSCAN, and choose between them for a dataset
- Choose a sensible number of clusters and evaluate clustering quality with silhouette and other metrics
- Reduce dimensionality with PCA and interpret explained variance and components
- Visualise high-dimensional data faithfully with t-SNE and UMAP and read the results honestly
Syllabus
Foundations & k-means Clustering
What unsupervised learning is, and when it helpsFree preview
k-means and choosing k
Density & Hierarchy — Beyond k-means
Hierarchical (agglomerative) clustering
DBSCAN — density-based clustering
Evaluating Clusters & PCA
Evaluating clustering quality
PCA for dimensionality reduction
Visualisation & Putting It Together
t-SNE and UMAP for visualisation
When unsupervised methods help — a full workflow
Lab — 10 Exercises & Solutions
Exercises 1–5
Exercises 6–10