Multiple linear regression: workflow, evaluation, and interpretation.
Use this session to move from the math of one line to a practical ML workflow with multiple features, model evaluation, assumptions, and business interpretation.
Session roadmap
Evaluation metrics
MAE, MSE, RMSE, R2, adjusted R2, baseline comparison, and interpreting model quality.
2Multiple regression
Many input features, coefficient interpretation, advertising dataset framing, polynomial feature teaser, and matrix view.
3Assumptions and risks
Linearity, independent errors, constant variance, normal residuals, multicollinearity, scaling, and regularization teaser.
Class outcome
By the end of Session 2, students should be able to build, evaluate, and explain a multiple linear regression model on a real dataset such as advertising spend vs sales.
End-of-session check
If a model has high R2 but unstable coefficients, what issue should we suspect?
Takeaway: multicollinearity or redundant features may be making coefficient interpretation unreliable.