Module · How a Model is Fitted✅Knowledge checkCheck: what is the model actually scoring?4 questions · Pass at 60%1. Squared error and absolute error are fitted to the same 200 markets, and they return different slopes. Which line is the better one?The squared-error line — it is the standard and has a formulaThe absolute-error line — it is less disturbed by unusual marketsNeither answer is available from the data; each cost ranks its own line firstWhichever scores lower under squared error, since that is the usual measure2. One market that spent almost nothing on TV and sold a great deal is added to the file. Which line moves further, and why?The absolute-error line, because it has no formula and must searchThe squared-error line, because a gap of 18 counts as roughly 340 once squaredBoth move the same amount — one row out of 201 is one rowNeither moves; a single row cannot change a fitted line3. You are forecasting stock for one warehouse, where being 500 units short once is far worse than being 20 units off twenty times. Which cost fits that?Absolute error, because every unit of error costs the sameSquared error, because it makes one large miss dominate the fitEither — the cost affects the score but not the modelNeither; this is a decision about the form, not the cost4. A cost function and an objective function differ how?A cost is for regression and an objective is for classificationThey are the same idea; an objective is sometimes made large rather than smallA cost is measured on training data and an objective on held-out dataAn objective has a formula and a cost does notCheck answers← Back to How a Model is FittedNext topic →
