In this lesson
- It starts with a question. You have 500 customer records and no idea how they divide. What are you doing when you group them?
- The idea you will find. Clusters are suggestions, not facts. — Change the settings and the groups change. A cluster is a hypothesis about structure that a person must then check against the world. Classification is judged differently — against labels you already trust, on data the model has never seen.
- How you learn it. By doing: sorting.
- Length. 5 short steps. Progress saves on this phone.
Part of Grouping & Labelling Data
Two different jobs that beginners mix up — and why the difference matters.