Skip to content

FRM Part I · FRM Exam Part I · Machine Learning and Prediction

A model predicts loan default using 40 predictors and achieves near-zero error on the training sample but much higher error on a separate validation sample. Which diagnosis and remedy is most appropriate?

The pattern of near-zero training error and much higher validation error indicates overfitting. The suitable remedy is to increase regularization or simplify the model so it generalizes better. Adding predictors or discarding validation data would worsen or conceal the problem.

  1. AUnderfitting; add more predictors and reduce regularization
  2. BOverfitting; increase regularization or simplify the modelCorrect
  3. CUnderfitting; shrink the training sample
  4. DOverfitting; remove the validation sample and rely on training error

Explanation

A large gap between low training error and high validation error signals overfitting: the model captures noise. Regularization or a simpler model reduces variance. Adding predictors would worsen the problem, and dropping validation hides it.

Did you get it right without looking?

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

Practise Machine Learning and PredictionFree diagnostic mock

More Machine Learning and Prediction questions