Math Atom · M-D12

Classification and discrimination metrics

10 minMath AtomStatistics

By the end you will be able to

  1. 1Build the confusion matrix at a cut-off and compute the true and false positive rates, precision and recall.
  2. 2Draw the ROC curve, and read the AUC as the probability that a random bad scores below a random good; convert it to a Gini coefficient.
  3. 3Compute the accuracy ratio from a CAP curve and the KS statistic, and say how they relate to AUC.
  4. 4Explain why a model can rank borrowers well and still predict the wrong default rates.