Supervised Learning
Learning from labeled examples where the correct target is already known.
Teach a router from resolved service tickets
Imagine a help desk that wants to route new requests to the right team.
- Input
An earlier ticket reads “The lift on floor two is stuck.” The words and location are information available when a ticket arrives.
- Label
A dispatcher confirmed that the facilities team handled the ticket. That team name supplies the training target.
- Prediction
For a new ticket, the model proposes a team from the message. A reviewer can later compare it with the team that actually resolved the request.
Choose labels the model could usefully predict
A historical team name can reflect an old routing rule or an accidental transfer. Check ambiguous labels before using them as examples. Also exclude fields filled in after resolution from the inputs. Otherwise the model could rely on information unavailable when a new ticket arrives.
Follow the idea
Training set describes the examples used to fit the router.
Unsupervised learning explores structure when the routing labels are unavailable.
Related concepts
Examples in the AI guides
Source · Google’s introduction to supervised learning
Return to Supervised Learning in the A–Z glossary