Regressor

Regressor is a term used in statistics and machine learning to describe an algorithm, model, or tool that predicts a continuous numerical value. In regression tasks, the goal is to learn the functional relationship between input features (predictors) and a continuous target variable.

Common regressor families include linear regression and its regularized variants (ridge, lasso, elastic net); nonlinear regression with polynomial features or kernel methods; and tree-based or ensemble approaches such as regression trees, random forest regression, gradient boosting regressor, and modern neural networks configured for regression.

In software libraries, a regressor often refers to an estimator object with fit and predict methods. In scikit-learn, for example, many regressors implement the RegressorMixin interface, and users evaluate performance on held-out data.

Evaluation typically uses metrics such as mean squared error, root mean squared error, mean absolute error, and R-squared. Training usually involves minimizing a loss function (often least squares) on labeled data. Regularization helps prevent overfitting, and cross-validation assesses generalization.

Regressors are applied in diverse domains, including finance, engineering, environmental science, and social sciences, for tasks such as forecasting, trend estimation, and calibration. The term regressor can also refer to an individual predictor variable in a regression model, i.e., an independent variable used to explain the target.

See also regression analysis, predictive modeling, and regression estimators.