A public-health clinic labels prior outbreak cases so a model can predict high-risk ZIP codes for outreach. Which training approach does that labeling enable?

Labeling prior outbreak cases so a model can predict high-risk ZIP codes enables supervised learning: the model trains on municipal examples where inputs (features about place, time, demographics, or case history) are paired with known outcomes or risk labels, then generalizes to score new ZIP codes for outreach. That approach works because outreach prioritization is a prediction problem with historical ground truth the clinic can curate. Unsupervised learning that trains on labeled municipal examples to predict risk fails by definition because unsupervised learning works without labeled outcomes to find structure, not by training on labeled examples to map features to known risk labels. Reinforcement learning that trains on labeled municipal examples to predict risk fails by definition because reinforcement learning learns from trial-and-error reward feedback on actions, not from labeled municipal examples paired with outcomes to predict static risk. Self-supervised learning that trains on labeled municipal examples to predict risk fails by definition because self-supervised learning creates its own training signals from unlabeled data rather than training on the clinic's curated labeled outbreak cases. Exam caveat: supervised risk scores can encode historical inequities in testing or reporting; labels must be audited for bias before they drive public-health resource allocation. Operational check: define label criteria with epidemiologists, hold out recent outbreaks for validation, track precision by ZIP subgroup, and require human review before automated outreach lists go live.