linear regression

Linear Regression

A statistical model that estimates a continuous outcome from one or more input variables by fitting the best straight line (or hyperplane) through the data.

When you'd see it: Revenue forecasting, pricing models, marketing mix models, compensation analysis, and any situation where you want to quantify 'how much does Y change when X changes by 1 unit.' The output is a coefficient for each predictor and an intercept.

Why it matters: Linear regression is the foundational predictive model — it's interpretable, computationally cheap, and tells you the direction and magnitude of each variable's relationship to the outcome. More complex ML models are often evaluated against a linear regression baseline, because if the linear model performs comparably, the additional complexity isn't worth it.

Common mistakes: Using linear regression for a binary outcome (churn/no churn, conversion/no conversion). When the outcome is categorical, logistic regression or a classifier is appropriate. Linear regression on a binary outcome can produce predicted probabilities outside [0,1], which are mathematically nonsensical.

Related logistic regression, R², RMSE, overfitting, correlation