Skip to content
Financial Machine Learning / Classification and Generalized Linear Models

Evaluating Classifiers: Precision, Recall, and ROC/AUC

Exercise 4 of the perceptron lesson flagged the trap: with 90% of ratings actions being "hold," a classifier that always predicts "hold" scores 90% accuracy while being completely useless. This note gives the vocabulary and tools to see past that trap — the confusion matrix, precision/recall, and the ROC curve — which is exactly how a credit-default classifier's threshold, not just its fitted parameters, becomes a business decision.

In this lesson

  1. Learning objectives
  2. The confusion matrix
  3. The metric vocabulary
  4. The threshold trade-off
  5. The ROC curve and AUC
  6. Worked example

Before you start

The full lesson, with worked examples and exercise solutions, is for members.

Read the full lesson

New to SelfMadeQuant.com? See membership options or browse the Financial Machine Learning syllabus.