Advanced Credit Risk Modeling for Basel/IFRS 9 using R/Python/SAS

📅 September 10th-11th, 2020 (9am-5pm)
🌍 English

About This Course

In this course, students learn how to do advanced credit risk modeling. We start by reviewing the Basel and IFRS 9 regulation. We then discuss how to leverage alternative data sources for credit risk modeling and do feature engineering. This is followed by an overview of variable selection and profit driven performance evaluation. We discuss some advanced modeling methods such as ensemble methods, neural networks, and Bayesian networks. We then cover low default portfolios and validation. The course concludes by reviewing stress testing. Modeling methods, performance measurement and benchmarks are discussed into great detail. The course provides a sound mix of both theoretical and technical insights, as well as practical implementation details. These are illustrated by several real-life case studies and examples. The course also features code examples in R, Python and SAS. Throughout the course, the instructors also extenisvely report upon their research and industry experience.

Course Outline

• Introduction
  • Instructor Team
  • Our Credit Risk Publications
  • Course Outline
  • Software
  • R/Python tutorials
  • Data sets
  • Disclaimer
• Credit Risk Modeling for Basel and IFRS 9
• Regulatory versus Economic Capital
• Basel I and II
• Basel III
• Basel IV
• Basel Approaches to Model Credit Risk
• Standardized Approach
• IRB Approach
• Default Rating Specifics
• Basel IRB Model Architecture
• Merton Model
• Risk Weight Functions
• Risk Weight Functions in R
• Risk Weight Functions in Python
• IFRS 9
• Credit Risk Analytics
• Alternative Data Sources for Credit Risk Modeling
  • Call Detail Record Data
  • Social Media Data
  • Clickstream Data
  • Google Street View
  • Google Trends
  • API/Web Scraping
  • Web Scraping in Python
  • Open Data
• Feature Engineering
  • Feature Engineering Defined
  • RFM features
  • Trend features
  • Ratio features
  • Logarithmic transformation
  • Power transformation
  • Box-Cox transformation
  • Yeo-Johnson transformation
  • Principal Component Analysis
  • Percentile coding
  • Thermometer coding
  • Performance Optimization
  • Feature engineering in R
  • Feature engineering in Python
• Variable Selection
  • Filter methods (gain, Cramer’s V, Fisher score)
  • Filters in R
  • Filters in Python
  • Forward/Backward/Stepwise regression
  • Forward/Backward/Stepwise in R
  • BART: Backward Regression Trimming
  • BART variable selection in R
  • Criteria for variable selection
• Profit Driven Performance Evaluation
  • Profit Driven Performance evaluation
  • Total and Average Misclassification Cost
  • Cutoff point tuning
  • MP and EMP
• Ensemble Methods
  • Bootstrapping
  • Bagging
  • Boosting
  • Random Forests
  • Random Forests in R
  • Random Forests in Python
  • XGBoost
  • XGBoost in R
  • XGBoost in Python
• Neural Networks
  • Neural Networks
  • Multilayer Perceptron
  • Deep Learning Neural Networks
  • Deep Learning for Credit Scoring: Do or Don't?
  • Neural Networks in Credit Risk
  • Neural Networks in R
  • Neural Networks in Python
  • Opening Neural Network Black Box
  • Variable Selection
  • Rule Extraction
  • Decompositional Rule Extraction
  • Pedagogical Rule Extraction
  • Quality of Extracted Rule Set
  • Rule Extraction Example
  • Two-Stage Model
  • Self-Organizing Maps
  • Self-Organizing Maps Example
  • Self-Organizing Maps Evaluated
  • SOMs in R
• Bayesian Networks
  • Bayesian Networks
  • Example Bayesian Network Classifier
  • Naive Bayes Classifier
  • Tree Augmented Naive Bayes Classifiers
  • Bayesian Networks in R
  • Bayesian Networks in Python
• Low Default Portfolios
  • Low Default Portfolios
  • Modeling Approaches
  • Shadow Rating Approach
  • Cumulative Logistic Regression in R
  • Cumulative Logistic Regression in Python
  • Sampling Approaches
  • Undersampling
  • Oversampling
  • Synthetic Minority Oversampling technique (SMOTE)
  • Adjusting Posterior Probability estimates
  • Adjusting Posterior Probability estimates in R
  • Adjusting Posterior Probability estimates in Python
  • Calibration Approaches
  • Calibration Approach in R
  • Calibration Approach in Python
  • EAD and LGD for Low Default Portfolios
• Model Validation
  • A regulatory perspective
  • Validation Overview
  • Developing a Validation Framework
  • Quantitative Validation
  • Backtesting PD
  • Backtesting PD at Level 0
  • System Stability Index in R
  • System Stability Index in Python
  • Backtesting PD at Level 1
  • ROC curve in R
  • ROC Curve in Python
  • Overrides
  • Backtesting PD at Level 2
  • Brier Score
  • Binomial Test
  • Binomial test in R
  • Binomial test in Python
  • Hosmer-Lemeshow Test
  • Normal Test
  • Vasicek One-Factor test
  • Vasicek One-Factor test in R
  • Vasicek One-Factor test in Python
  • Data Aggregation
  • Risk Rating Philosophy
  • Example Traffic Light Indicator Dashboard
  • Example Action Scheme
  • Backtesting LGD and EAD (Bootstrapping, F-test, Student’s t test, Wilcoxon signed rank test, loss shortfall, mean absolute deviation)
  • Benchmarking
  • Rating Difference Histogram
  • Spearman’s Rank-Order Correlation
  • Kendall’s Tau
  • Goodman-Kruskal Gamma
  • Benchmarking Example
  • Ranking statistics in R
  • Ranking statistics in Python
  • Qualitative Validation
  • Use test
  • Data quality
  • Model design
  • Documentation
  • Corporate governance and management oversight
  • Privacy and Security
• Stress Testing
  • Stress Testing
  • Stress Testing Applications
  • Regulatory perspective
  • Challenges in stress testing
  • Types of stress testing
  • Sensitivity-based stress testing analysis
  • Scenario analysis
  • Historical scenarios
  • Hypothetical scenarios
  • Pillar 1 versus Pillar 2 Stress Testing
  • Macroeconomic approaches to stress testing
  • Stress testing governance