Linear Regression

Also known as · regression

A model that predicts a continuous target as a weighted sum of the features plus an intercept, y^=β0+β1x1+⋯+βkxk\hat{y} = \beta_0 + \beta_1 x_1 + \dots + \beta_k x_k, fitted by OLS.

When to use

Use it as the baseline for any "how much?" question and evaluate with MAE, RMSE and R-Squared on the test set. A coefficient is the change in prediction per unit of a feature, holding the others constant: association, not causation.

See DS Lec 4.

Appears in