Prerequisites
Probability, regression, accounting basics and a careful understanding of the outcome definition.
What to study
Study probability of default, loss given default and exposure at default. Distinguish a predictive score from a calibrated probability and a realised loss from an expected loss.
A project to complete
Create a time-split default model on a dataset with documented definitions. Compare calibration and discrimination, check missingness and explain which observations are available at the decision time.
A useful deliverable
Publish the problem definition, documented data, assumptions, reproducible code, a simple comparison and a short discussion of failure cases. Label demonstrations and provisional findings clearly.
Continue reading
Use the first supporting lesson, the next practical topic, and the source reading lists. The repository map shows where earlier material is held.