Regression (Machine Learning)

Regression is a form of supervised learning in which a model learns the relationship between input features and a continuous numeric target, then predicts that number for new data. Linear regression fits a straight line or plane; other methods fit curves or more complex shapes.

Why it matters

Regression answers "how much" and "how many" questions: forecasting demand, estimating delivery time, predicting a page's load time from its size and request count. Common error measures are mean absolute error and root mean squared error, which are in the same units as the target. A model that fits the training data tightly may still perform poorly on new data, so evaluate on held-out samples.

Example

Simple linear regression:
response_ms = 80 + 0.4 * page_kb

Page of 500 KB -> predicted 80 + 200 = 280 ms

The intercept (80) and slope (0.4) are the parameters learned from historical measurements.

Common mistakes

• Extrapolating far beyond the range of the training data.
• Reading a strong fit as proof of cause. See correlation versus causation.
• Calling logistic regression a numeric predictor. Despite its name it is used for classification.

Related

• Classification
• Supervised learning
• Overfitting
• Correlation vs causation

This page is general information written by our editorial team and updated when documentation changes. Servers and providers differ, so test changes safely first. See our editorial policy and disclaimer.