Unsupervised learning
Discovering structure in unlabeled data: clustering, dimensionality reduction and anomaly detection.
Course Duration: 6h
What You'll Learn
- Segment a population with no prior labels
- Choose the number of groups using defensible criteria
- Reduce dimensionality without destroying the useful signal
- Visualize high-dimensional data honestly
- Spot anomalies in a stream of measurements
Prerequisites
- Python and pandas
- Notions of linear algebra
Course Modules
- What we look for when there are no labels
- k-means: principle, initialization and limits
- Choosing the number of clusters: elbow and silhouette score
- Hierarchical clustering and dendrograms
- DBSCAN and clusters of arbitrary shape
- Principal component analysis
- t-SNE and UMAP: reading projections with care
- Anomaly detection: isolation forest and statistical methods
- Gaussian mixture models
- Project: a documented customer segmentation
Assessment and certificate
The course ends with a 40-question exam covering every module. On success, a certificate of completion is issued; its number is verifiable on the platform.
Free courses, by contrast, end with a 5-question quiz and a preview of the certificate, without certification.