You have unlabeled documents and want to group similar ones. What fits?
Why C is the answer
Clustering
Clustering needs no labels. Regression and classification do.
In detail
- Clustering groups data with no labels.
- A person names the groups later.
Use case
An illustrative scenario, not a real incident.
A legal team groups thousands of unlabeled contracts by similarity, with clustering.
See it in different ways
mindmap
root((ML basics))
Labeled data
Classification
Regression
Unlabeled data
Clustering
Validation
Cross validation
Held-out test setflowchart TD
Q{"What do you have?"}
Q -->|"No labels"| A0["Clustering"]
Q -->|"Labels"| A1["Classification or regression"]Key point: Clustering needs no labels.
Why the other options are wrong
- A. Binary classification with a labeled training set
- Classification needs labels.
- B. Reinforcement learning with a reward signal
- That needs a reward.
- D. Linear regression on a labeled target value
- Regression needs a labeled target.