Skip to content

FRM Part I · FRM Exam Part I · Machine-Learning Methods

A data scientist estimates a credit-scoring model with 200 candidate predictors, believing only about 15 truly matter, and wants the fitted model itself to identify and discard the irrelevant ones. Which approach is most appropriate, and why?

LASSO is the appropriate choice. Its absolute-value (L1) penalty can set coefficients of irrelevant predictors exactly to zero, giving a sparse model that selects variables automatically. Ridge only shrinks coefficients toward zero and keeps every predictor, while OLS keeps all of them too.

  1. ARidge regression, because the squared penalty drives irrelevant coefficients exactly to zero
  2. BLASSO regression, because the absolute-value penalty can set irrelevant coefficients exactly to zeroCorrect
  3. COrdinary least squares, because it is unbiased and so ignores irrelevant predictors
  4. DRidge regression, because it increases lambda until all coefficients equal their OLS values

Explanation

The L1 penalty in LASSO has a corner at zero, so as lambda grows some coefficients become exactly zero, yielding a sparse model. Ridge only shrinks coefficients and keeps all predictors. OLS retains every predictor and has high variance with many predictors.

Did you get it right without looking?

One question tells you little. A timed set on Machine-Learning Methods shows your real accuracy, how long you take and where you lose marks.

Practise Machine-Learning MethodsFree diagnostic mock

More Machine-Learning Methods questions