Skip to content

FRM Part II · FRM Exam Part II · Advances in Artificial Intelligence: Implications for Capital Markets Activities

A credit desk uses a neural network trained on 2015-2019 market data to flag stressed issuers. In 2022, following a sharp shift in rates and inflation, its precision falls from 80% to 50% although the code and features are unchanged. The model had 200 flags in 2022. Assuming the precision drop is accurate, how many fewer true positives were there among the 200 flags than the 80% precision would imply, and what is the most likely cause?

There were 60 fewer true positives: 80% of 200 is 160 while 50% of 200 is 100. The most likely cause is concept drift, because the relationship between features and defaults changed after the rate and inflation regime shift, while the model was trained on the earlier period.

  1. A60 fewer; concept drift, as the relationship between features and outcomes changedCorrect
  2. B60 fewer; overfitting to the 2022 data
  3. C100 fewer; concept drift, as the relationship between features and outcomes changed
  4. D30 fewer; label leakage during training

Explanation

Expected true positives at 80% precision are 160; at 50% they are 100, a shortfall of 60. Unchanged code and features with a changed macro regime points to concept drift, where the data-generating relationship shifts. Overfitting to 2022 is wrong because the model was trained on 2015-2019. Leakage would inflate, not deflate, performance.

Did you get it right without looking?

One question tells you little. A timed set on Advances in Artificial Intelligence: Implications for Capital Markets Activities shows your real accuracy, how long you take and where you lose marks.

Practise Advances in Artificial Intelligence: Implications for Capital Markets ActivitiesFree diagnostic mock

More Advances in Artificial Intelligence: Implications for Capital Markets Activities questions