Classification vs. regression

In classification the target is a category: spam / not spam, digit 0–9, disease A / B / C. The model outputs a class (often with a probability for each class). Binary classification has two classes; multiclass has more.

In regression the target is a continuous number: a price, a temperature, a time until failure. The model outputs a real value and is judged by how far off it is (RMSE, MAE).

The distinction drives the choice of loss and metrics: squared error for regression, cross-entropy for classification; RMSE and R² versus accuracy, precision and recall.