Machine Learning BasicsLesson 3
Lesson 38 min
Unsupervised Learning
Finding structure in data that has no labels.
What you will learn
- ✓What unsupervised learning is for
- ✓Clustering with K-Means
- ✓Dimensionality reduction with PCA
Explanation
Unsupervised learning works on data with no answers attached — the goal is to discover hidden structure.
Clustering (e.g. K-Means) groups similar items together, like segmenting customers by behaviour. Dimensionality reduction (e.g. PCA) compresses many features into a few, useful for visualisation and speed.
Because there is no label, you judge results by usefulness, not accuracy.
Real-world use
Retailers cluster shoppers into segments (bargain hunters, loyalists, etc.) to target promotions — without anyone labelling the groups in advance.
Common mistakes
- • Expecting a single 'correct' answer — unsupervised results are interpretations, not exact truths.
Practice
Describe a dataset you could cluster and what useful groups you'd hope to find.
Knowledge check
0/1 answered1. K-Means is an example of...
Answer all questions to check.