Zoom Training Details:

Day 1:

Duration	Topic
80 min	Data Pre-processing and COVID-19 Credit scoring techniques for corporate & retail data Outlier detection Standardisation & categorisation Default definitions Non-monotone relations between defaults and risk factors Weight-of-evidence, categorisation, splines Multivariate non-monotone relationship COVID-19 challenges
20 min	Break
80 min	Machine Learning Concepts Statistical and machine learning thinking Workflow in machine learning Loss functions Train-test validation split Standardisation and scaling Bias-variance trade-off Cross-validation Hyper-parameters and tuning Supervised and unsupervised learning Classification and regression vs. dimensionality reduction

Day 2:

Duration	Topic
80 min	Probabilities of Default (PD)– Expected PDs Discrete time models (Logit & Probit) Comprehensive modelling including vintage, age and time effects Maximum likelihood estimation Significance testing Prediction
20 min	Break
80 min	Machine Learning Validation Validation framework Backtesting PDs: discrimination, calibration, stability Confusion matrix, sensitivity, specificity, accuracy, ROC curves, AUC & accuracy ratio Brier scores Binomial test Jeffreys prior test Hosmer-Lemeshow Calibration curve and reliability diagram Tests on model stability Portfolio dependence Traffic lights Qualitative validation

Day 3:

Duration	Topic
80 min	Probabilities of Default (PD) – Crisis PDs Model stability, discrimination & calibration Through-the-cycle & point-in-time Predicting Financial Crises Margin of Conservatism (MoC) CECL and IFRS 9 models for multi-period PDs
20 min	Break
80 min	Unsupervised Machine Learning for PDs K Means clustering Hierarchical clustering Principal component analysis

Day 4:

Duration	Topic
80 min	Supervised Machine Learning for PDs Logistic regression K nearest neighbours Decision trees Bagging and boosting Random forests
20 min	Break
80 min	Supervised Machine Learning for PDs Neural networks Deep learning

---

Back to Zoom Training