Regression analysis
Regression analysis models how an outcome relates to one or more explanatory variables. In machine learning, regression usually predicts numeric targets; the broader statistical family also includes models for categorical outcomes, such as logistic regression.
A product team could fit monthly revenue against advertising spend and seasonal indicators. A linear model assigns coefficients to those inputs and uses them to predict revenue for another month. Inspecting prediction errors helps identify where the model misses patterns or where its assumptions are unsuitable.
A positive spend coefficient does not prove extra advertising caused the revenue increase. Spend may rise when demand is already high, or alongside other changes. Predictive association and causal effect are different claims. Correlated inputs can also make individual coefficients difficult to interpret.
Numeric prediction contrasts with classification, which assigns categories. Logistic regression belongs to the statistical regression family but models class probabilities and is commonly used as a classifier. State the target and model form rather than relying on the name. For forecasting, evaluate on later periods and consider whether the historical relationship can persist.
Sources
- scikit-learn: Linear models — Documents ordinary least squares, regularized regression and logistic regression as a classification model.
Go deeper
- Google: Linear regression course
Work through weights, bias and numeric predictions before interpreting a fitted model.