Broad Categories of Machine-Learning Models: Supervised Learning Models — Part d

Article 2d of 25 — Regression & Evaluation Metrics

1.1.3 Regression

Regression is a supervised-learning approach used to predict numerical outcomes from one or more input features. It helps quantify relationships, estimate future values, and support planning in applications such as forecasting demand, calculating costs, predicting delivery times, and estimating remaining useful life. Effective regression requires a clearly defined target, representative data, an appropriate model and loss function, and evaluation measures that reflect the practical consequences of prediction errors.

Figure 1.1.3. Regression maps features to continuous predictions, with loss choice, diagnostics, uncertainty, and evaluation defining model quality.

Regression models predict a continuous numerical value.

Technically, a regression model learns a function that maps a feature vector to a value on a numerical scale. The predicted value may represent a conditional mean, median, quantile, rate, duration, cost, or another continuous quantity, depending on…