Zoom Training Details:

Day 1:

Duration	Topic
80 min	PD/LGD principles Illiquidity and leverage External Data Default Rate Cycles Non-linear relations between defaults and risk factors Linear, Logit & Probit regressions for PD Linear & Fractional Response regressions for LGD Predictability and pain points Transitional (climate and geopolitical) risks Climate change challenges Multi-period forecasting Stress-testing Impact of COVID-19, income and payment shocks on credit risk COVID-19 challenges Interest and expense increases
20 min	Break with industry speaker
80 min	You will join a breakout in Python or R language to apply: Key Programming Concepts Feature engineering Outcome engineering: loan amortization, default and workout LGDs Time-vintage-age (TVA) analysis analysis Visualisation and Reporting through Dashboards Efficient presentation Dashboards and user interfaces Communication with non-modelers

Day 2:

Duration	Topic
80 min	Machine Learning Concepts for Probabilities of Default (PDs) Statistical and machine learning thinking Workflow in machine learning Loss functions Regularisation penalty Learning rate Train-test validation split Standardisation and scaling Bias-variance trade-off Machine Learning Validation for PDs Validation framework Backtesting PDs: discrimination, calibration, stability Confusion matrix, sensitivity, specificity, accuracy, ROC curves, AUC & accuracy ratio Brier scores Calibration tests (Binomial test, Jeffreys prior test, Hosmer-Lemeshow) Tests on model stability Portfolio dependence Traffic lights Qualitative validation
20 min	Break with industry speaker
80 min	You will join a breakout in Python or R language to apply: Estimate a comprehensive PD model Illiquidity and leverage Age Macroeconomy Validate a comprehensive PD model Validation function: metrics, real-fit, PD distributons and calibration curve and reliability diagram Discrimination Calibration tests and calibration curve Model stability

Day 3:

Duration	Topic
80 min	Crisis PDs Asymptotic Single Risk Factor Model (ASRF) Asset correlations Basel calibrations Stress testing Model risk Downturn LGD based on Workout Recoveries Discount factors Resolution Bias OCC proposal Multi-period PDs and Lifetime PDs for CECL and IFRS 9 Weighted Average Life based Framework Constituents: PDs, survival probabilities, expected losses, discount factors
20 min	Break with industry speaker
80 min	You will join a breakout in Python or R language to apply: Prediction of Crisis PDs Through-the-cycle vs. point-in-time Predicting financial crises Stress-testing for economic downturns Assessment of model risk Margin of Conservatism Expected loss modeling for IFRS 9 Macroeconomic forecasts Multi-period PD forecasts based on the macroeconomy and lifecycle Prediction of 12 month and lifetime expected losses Significant increase in credit risk (SICR) Unexpected loss modeling for capital ASRF application to loss distributions Value-at-Risk Expected Shortfall

Day 4:

Duration	Topic
80 min	Supervised Machine Learning for PDs Bagging and boosting Network architecture Activation function Deep learning Explainable AI Shapley values Ale plots Linear and non-linear feature importance
20 min	Break with industry speaker
80 min	You will join a breakout in Python or R language to apply: Fit Machine Learning Models Random forests Light Gradient Boosting Machines (GBM) Extreme Gradient Boosting Machines (XGBM) Cross-validation Hyper-parameters and tuning Do it like Google: Introduction into Keras/Tenserflow Artificial neural network

---

Back to Live