Supervised Learning
Supervised learning fits a model using examples paired with target outputs. Targets can be categories for classification or numerical values for regression; the learning procedure uses them to assess and improve predictions.
A support team can label past tickets as billing, login or outage, then train a classifier to predict those labels from ticket text. The model learns a mapping from inputs to targets rather than receiving a hand-written rule for every message.
Targets need not be perfect. Two reviewers may disagree about a mixed billing-and-login ticket, and the chosen labeling rule determines what training rewards. Check label quality and measure performance for the categories the product actually needs.
Evaluate on held-out examples instead of treating training accuracy as evidence of useful deployment. Keep related or near-duplicate tickets together when splitting the data. Supervised fine-tuning continues this type of training from a pretrained model. Showing examples in a prompt instead changes the input at inference time; it does not by itself fit new parameters. The distinction matters when deciding whether to change data, train weights or improve the prompt.
Sources
- Google: Supervised learning — Explains labeled examples, fitting, classification and regression.
Go deeper
- scikit-learn: Getting started docs
Fit a model with input rows and targets, predict new outputs and build an evaluation pipeline.