Regression predicts a number, such as a price. Classification predicts a category: spam or not spam, fraud or fine, cat or dog. The model still learns from examples, but every example in the Dataset carries a label that was put there by a person or an earlier system.
Our running case is a spam filter. We have 100 emails, each labelled. The Trainer studies them and produces a Model. When a new email arrives, the Model answers with a category.
💡Good to know
Ask one question first: is the answer a number on a scale, or one of a short list of labels?