electrical-connection-energy-infrastructure-supply-distribution-of-energy-transmission-hig

IFRS 9: Credit Risk Modeling II

Intensive course on IFRS 9 credit risk methodologies applying traditional econometric and machine learning models as well as innovative probabilistic machine learning and quantum computing models.

COURSE OBJECTIVE

The COVID-19 pandemic emerged just two years after the 2018 implementation of IFRS 9. The pandemic stressed and affected the predictive power of the models and methodologies, posing significant challenges for creating provisions for impaired assets. In the wake of the pandemic shock, subsequent regulatory and government actions, as well as the recent unprecedented set of risk events such as war, European energy supply insecurity and global inflationary pressures, banks have gradually planned recalibrate the IFRS 9 expected credit loss (ECL) models to improve their accuracy and incorporate lessons learned. However, although adjustments to the models are necessary, new macroeconomic shocks continue to appear, influenced by high uncertainty.

Entities have faced several challenges. The first was the significant increase in credit risk (SICR) that was based on inaccurate or incomplete information. Second, the probability of default (PD) was not sensitive enough to forward-looking and non-linear information. Third, banks applied overlays more frequently, but did not justify or quantify them.

Some prestigious consulting firms propose to automate more processes, develop challenging models of PDs and ECL expected credit losses.

Therefore, we have created a course with a greater number of lifetime PD estimation and calibration models, we have increased the artificial intelligence models and added models based on quantum algorithms that on the one hand can be challenging models of the traditional ones and that will help measure nonlinear relationships.

However, the core of the course is to explain in detail credit risk methodologies to estimate and calibrate the lifetime parameters of PD, LGD and EAD adjusted to the IFRS 9 standard using econometric models, Bayesian approach, traditional machine learning, quantum machine learning and quantum algorithms.

All models must quantify the uncertainty inherent in financial inferences and predictions to be useful in financial risk management and decision making. Model parameters and outputs can have a range of values with associated probabilities. Therefore, mathematically sound probabilistic models are needed that adapt to inaccuracies and that quantify uncertainties with logical consistency. Therefore, we have included probabilistic machine learning models, that is, machine learning algorithms together with probabilistic modeling and Bayesian decision theory. These algorithms offer modern and powerful solutions in today's complex financial and economic environment.

This course includes more than 12 methodologies and exercises to estimate PD Lifetime in retail, mortgage, SME and corporate portfolios, for example, the Exogenous Maturity Vintage EMV model, Markov models, survival models, transition matrices, Deep Learning, Monte Carlo simulation quantum algorithms among others.

Forecasting and stress testing methodologies have been incorporated to generate forward looking economic scenarios. Regarding the subject, there are several modules dedicated to the design of scenarios where the interaction between the macroeconomic variables and the Lifetime PD are exposed. In addition, stress testing methodologies for IFRS 9 credit risk provisions are explained.

Regarding the LGD Lifetime, machine learning models are shown to improve the accuracy of the parameters. And regarding the EAD Lifetime, vintage models for lines of credit are explained, as well as econometric models for prepayment.

A pricing tool is delivered, which includes the estimate of ECL 12m and ECL Lifetime, regulatory capital, Raroc and Hurdle rate.

Quantum Machine Learning is the integration of quantum algorithms within Machine Learning programs. Machine learning algorithms are used to compute vast amounts of data, quantum machine learning uses qubits and quantum operations or specialized quantum systems to improve the speed of computation and data storage performed by algorithms in a program. For example, some mathematical and numerical techniques from quantum physics are applicable to classical deep learning. A quantum neural network has computational capabilities to decrease the number of steps, the qubits used, and the computation time.

The objective of the course is to show the use of quantum computing and tensor networks to improve the calculation of machine learning algorithms.

We show how quantum algorithms speed up the calculation of Monte Carlo simulation, the most powerful tool for developing credit risk models, representing an important advantage for calculating economic capital, lifetime PD and creating stress testing scenarios.

The objective of the course is to expose classical models against quantum models, explain the scope, benefits and opportunities.

To facilitate learning, most macros are delivered in Jupyter Notebook, an interactive web environment for running R and Python code, which includes videos, images, formulas, etc. that help the analysis and explanation of the methodologies.

WHO SHOULD ATTEND?

This program is aimed at risk managers, analysts and consultants who are immersed in the development, validation or audit of IFRS 9 credit risk models or for all those interested. For a better understanding of the topics, it is recommended that the participant have knowledge of statistics.

Friday, December 1, 2023
Monday, March 4, 2024
Europe: Mon-Fri, CEST 16-19 h
America: Mon-Fri, CDT 18-21 h
Asia: Mon-Fri, IST 18-21 h
Price: 7.900 €
Duration: 40 h
Presentations in PDF

Exercises in Excel, R, Python y Jupyterlab

The recorded video of the 40-hour course is delivered.
yulia.tsedryk@fermacrisk.com
Banks
IFRS 9 PDF

Agenda

Module 0: Quantum Computing and Algorithms (Optional)

Future of quantum computing in banking

Is it necessary to know quantum mechanics?

QIS Hardware and Apps

quantum operations

Qubit representation

Measurement

Overlap

matrix multiplication

Qubit operations

Multiple Quantum Circuits

Entanglement

Deutsch Algorithm

Quantum Fourier transform and search algorithms

Hybrid quantum-classical algorithms

Quantum annealing, simulation and optimization of algorithms

Quantum machine learning algorithms

Exercise 1: Quantum operations

QUANTUM COMPUTING
Module 1: Exploratory Analysis

Exploratory Data Analysis EDA

data sources

Data review

Target definition

Time horizon of the target variable

Sampling

Random Sampling

Stratified Sampling

Rebalanced Sampling

Exploratory Analysis:

histograms

Q Q Plot

Moment analysis

boxplot

Treatment of Missing values

Multivariate Imputation Model

Advanced Outlier detection and treatment techniques

Univariate technique: winsorized and trimming

Multivariate Technique: Mahalanobis Distance

Over and Undersampling Techniques

Random oversampling

Synthetic minority oversampling technique (SMOTE)

Module 2: Feature engineering

Feature engineering

Data Standardization

Variable categorization

Equal Interval Binning

Equal Frequency Binning

Chi-Square Test

binary coding

WOE Coding

WOE Definition

Univariate Analysis with Target variable

Variable Selection

Treatment of Continuous Variables

Treatment of Categorical Variables

gini

Information Value

Optimization of continuous variables

Optimization of categorical variables

Exercise 2: EDA Exploratory Analysis

Exercise 3: Detection and treatment of Advanced Outliers

Exercise 4: Multivariate model of imputation of missing values

Exercise 5: Univariate analysis in percentiles in R

Exercise 6: Continuous variable optimal univariate analysis in Excel

CREDIT SCORING
Unsupervised Learning

Module 3: Unsupervised models

Hierarchical Clusters

K Means

standard algorithm

Euclidean distance

Principal Component Analysis (PCA)

Advanced PCA Visualization

Eigenvectors and Eigenvalues

Exercise 7: Segmentation of the data with K-Means R

Supervised Learning

Module 4: Support Vector Machine SVM

SVM with dummy variables

SVM

optimal hyperplane

Support Vectors

add costs

Advantages and disadvantages

SVM visualization

Tuning SVM

kernel trick

Exercise 8: Credit Scoring Support Vector Machine

Module 5: Ensemble Learning

set models

bagging

bagging trees

Random Forest

Boosting

adaboost

Gradient Boosting Trees

Advantages and disadvantages

Exercise 9: Credit Scoring Boosting in R

Exercise 10: Credit Scoring Bagging in R

Exercise 11: Credit Scoring Random Forest, R and Python

Exercise 12: Credit Scoring Gradient Boosting Trees

AI CREDIT SCORING
Module 6: Deep Learning Feed Forward Neural Networks

Single Layer Perceptron

Multiple Layer Perceptron

Neural network architectures

activation function

sigmoidal

Rectified linear unit (Relu)

The U

Selu

hyperbolic hypertangent

Softmax

other

Back propagation

Directional derivatives

gradients

Jacobians

Chain rule

Optimization and local and global minima

Exercise 14: Credit Scoring using Deep Learning Feed Forward

Module 7: Deep Learning Convolutional Neural Networks CNN

CNN for pictures

Design and architectures

convolution operation

descending gradient

filters

strider

padding

Subsampling

pooling

fully connected

Credit Scoring using CNN

Recent CNN studies applied to credit risk and scoring

Exercise 15: Credit scoring using deep learning CNN

Module 8: Deep Learning Recurrent Neural Networks RNN

Natural Language Processing

Natural Language Processing (NLP) text classification

Long Term Short Term Memory (LSTM)

hopfield

Bidirectional associative memory

descending gradient

Global optimization methods

RNN and LSTM for credit scoring

One-way and two-way models

Deep Bidirectional Transformers for Language Understanding

Exercise 16: Credit Scoring using Deep Learning LSTM

Module 9: Generative Adversarial Networks (GANs)

Generative Adversarial Networks (GANs)

Fundamental components of the GANs

GAN architectures

Bidirectional GAN

Training generative models

Credit Scoring using GANs

Exercise 17: Credit Scoring using GANs

Module 10: Calibration of Machine Learning and Deep Learning

Hyperparameterization

grid search

random search

Bayesian Optimization

Train test split ratio

Learning rate in optimization algorithms (e.g. gradient descent)

Selection of optimization algorithm (e.g., gradient descent, stochastic gradient descent, or Adam optimizer)

Activation function selection in a (nn) layer neural network (e.g. Sigmoid, ReLU, Tanh)

Selection of loss, cost and custom function

Number of hidden layers in an NN

Number of activation units in each layer

The drop-out rate in nn (dropout probability)

Number of iterations (epochs) in training a nn

Number of clusters in a clustering task

Kernel or filter size in convolutional layers

pooling size

batch size

Interpretation of the Shap model

Exercise 18: Optimization Credit Scoring Xboosting, Random forest and SVM

Exercise 19: Credit Scoring Optimized Deep Learning and Model Interpretation

Module 11: Construction of the Scorecard

scoring assignment

Scorecard Classification

Scorecard WOE

Binary Scorecard

Continuous Scorecard

Scorecard Rescaling

Factor and Offset Analysis

Scorecard WOE

Binary Scorecard

Reject Inference Techniques

cut-off

parceling

Fuzzy Augmentation

Machine Learning

Advanced Cut Point Techniques

Cut-off optimization using ROC curves

Exercise 20: Construction of Scorecard in Excel, R and Python

DEEP LEARNING
Module 12: Quantum Credit Scoring and PD

What is quantum machine learning?

Qubit and Quantum States

Quantum Automatic Machine Algorithms

quantum circuits

K means quantum

Support Vector Machine

Support Vector Quantum Machine

Variational quantum classifier

Training quantum machine learning models

Quantum Neural Networks

Quantum GAN

Quantum Boltzmann machines

Quantum machine learning in Credit Risk

Quantum machine learning in credit scoring

quantum software

Exercise 21: Quantum K-means

Exercise 22: Quantum Support Vector Machine to develop credit scoring model

Exercise 23: Quantum feed forward Neural Networks to develop a credit scoring model and PD estimation

Exercise 24: Quantum Convoluted Neural Networks to develop a credit scoring model and PD estimation

Module 14: Tensor Networks for Machine Learning

What are tensor networks?

Quantum Entanglement

Tensor networks in machine learning

Tensor networks in unsupervised models

Tensor networks in SVM

Tensor networks in NN

NN tensioning

Application of tensor networks in credit scoring models

Exercise 25: Construction of credit scoring and PD using tensor networks

QUANTUM MACHINE LEARNING
Module 15: Probabilistic Machine Learning

Probability

Gaussian models

Bayesian Statistics

Bayesian logistic regression

Kernel family

Gaussian processes

Gaussian processes for regression

Hidden Markov Model

Markov chain Monte Carlo (MCMC)

Metropolis Hastings algorithm

Machine Learning Probabilistic Model

Bayesian Boosting

Bayesian Neural Networks

Exercise 26: Gaussian process for regression

Exercise 27: Bayesian neural networks

PROBABILISTIC MACHINE LEARNING
Default Probability

Module 16: Probability of Default PD

PD estimation

econometric models

Machine Learning Models

Data requirement

Risk drivers and credit scoring criteria

Rating philosophy

Pool Treatment

PD Calibration

Default Definition

Long run average for PD

Technical defaults and technical default filters

Data requirement

One Year Default Rate Calculation

Long-Term Default Rate Calculation

PD Model Risk

Conservatism Margin

PD Calibration Techniques

Anchor Point Estimate

Mapping from Score to PD

Adjustment to the PD Economic Cycle

Rating Philosophy

PD Trough The Cycle (PD TTC) models

PD Point in Time PD (PD PIT ) models

PD Calibration of Models Using Machine and Deep Learning

Margin of Caution

Exercise 27: PD Calibration Models

Exercise 28: PD calibration in Machine Learning models

Exercise 29: Modeling the Margin of Caution PD

Module 17: Econometric and AI Models of PD

PD estimation

Treatment of Panel data

Econometric models to estimate PD

PD Logistic Regression

PD Probit Regression

PD COX regression of survival

PD Log-log Complementary

PD Regression Data Panel

PD Bayesian Logistic Regression

PS Regression Lasso

PS Neural Networks

PS Quantum Neural Networks

PD Calibration

Calibration of econometric models

Anchor Point Estimate

PD calibration by vintages or vintages

vintage analysis

PS Marginal

PS Forward

Cumulative PD

Exercise 30: Calibration of PD with COX regression

Exercise 31: Calibrating PD with logistic regression with panel data in R

Exercise 32: Calibration of PD with Neural Networks in R

Exercise 33: Calibrating PD with Bayesian Logistic Regression in Python

Exercise 34: Calibration of the PD LASSO regression

Exercise 35: PD Calibration with Quantum Neural Networks in Python

Module 18: PD Calibration

Concept of adjustment to central tendency

Bayesian approach

PD calibration in developed countries

PD calibration in emerging countries

Scaled PD Calibration

Scaled Likelihood ratio calibration

Smoothing of PD curves

quasi moment matching

approximation methods

Scaled beta distribution

Asymmetric Laplace distribution

rubber function

Platt scaling

Broken curve model

Isotonic regression

Gaussian Process Regression

Exercise 36: PD calibration using Platt scaling and isotonic regression

Exercise 37: PD calibration using Gaussian Process Regression

Exercise 38: Calibration of the PD asymmetric Laplace distribution

Module 19: Bayesian PD and Gaussian Process

Bayesian and deterministic approach

expert judgment

prior distributions

Bayes' theorem

posterior distributions

Bayesian PD Estimation

Markov Chain–Monte Carlo MCMC approach

credibility intervals

Bayesian PD in practice

Calibration with Bayesian approach

Process Gaussian regression

Exercise 39: Logistic Model Bayesian PD in Python

Exercise 40: PD using MCMC in R

Exercise 41: PD using Process Gaussian Regression

Module 20: Low Default Portfolio PD (PD LDP)

Confidence interval approach for PD LDP

PD estimation without correlations

PD estimation with correlations

One-period and multi-period estimation

Bayesian PD estimation for LDP

Neutral Bayesian

Conservative Bayesian

expert judgment

Real analysis of PD of Corporate, Sovereign and Retail portfolios

LASSO regression to measure corporate default rate

Exercise 42: PD LDP confidence interval approach in R

Exercise 43: Multiperiod confidence interval approach PD LDP

Exercise 44: Neutral Bayesian PD in R

Exercise 45: Conservative Bayesian PD in R

IFRS 9 : EXPECTED CREDIT LOSSES
Module 21: Transition Matrices and Temporary Structure of PD

Temporary structure of PD in IFRS 9

Properties of transition matrices

Markov chains

Multi-year transition matrix

discrete time

continuous time

Generating Matrix

Exponential of a matrix

duration method

Cohort method

error management

PD temporary structure

Calibration of the temporal structure of the PD

Levenberg–Marquardt Algorithm

Economic Cycles

Calibration of the temporal structure of the PD for LDP

Exercise 46: Analysis and error exercise of Transition Matrix using cohort and duration approach in Python

Exercise 47: Calibration of the temporal structure of the PD

Module 22: Lifetime PD Models

PD Lifetime consumer portfolio

PD Lifetime mortgage portfolio

PD Lifetime Wallet Credit Card

PD Lifetime portfolio SMEs

vintage model

Exogenous Maturity Vintage EMV Model

decomposition analysis

Advantages and disadvantages

Basel ASRF model

Matrix ASRF model

Leveraging IRB in IFRS 9

Advantages and disadvantages

Regression Models

Logistic Multinomial Regression

Ordinal Probit Regression

Survival Models

Kaplan–Meier

Cox regression

Advantages and disadvantages

Markov models

Multi-State Markov Model

Cox Semiparametric Model

Advantages and disadvantages

Machine Learning Model

SVM: Kernel Function Definition

Neural Network: definition of hyperparameters and activation function

Quantum Machine Learning Models

PD Lifetime Extrapolation Models

Exercise 48: PD Lifetime using vintage EMV Decomposition model

Exercise 49: PD Lifetime using multinomial regression in R

Exercise 50: PD Lifetime using Markov model

Exercise 51: PD Lifetime using matrix ASRF model

Exercise 52: PD Lifetime using SVM in Python

Exercise 53: PD Lifetime using Neural Network in Python

Exercise 54: PD Lifetime using Quantum Neural Network in Python

Exercise 55: PD Lifetime using Quantum SVM in Python

Module 23: Advanced Calibration Lifetime PS

Calibration by nonlinear equation systems

Temporal structure calibration with Vasicek model

Calibration of transition matrices using Vasicek

Bayesian calibration

Fitting curve distributions

Extrapolation Calibration

Gamma Adjustment

exponential accelerator

Lifetime PD Recalibration

Nelson Siegel Calibration

Exercise 56: Lifetime Advanced Calibration Models PD Nelson Siegel

Exercise 57: Lifetime PD Advanced Calibration Models Gamma Adjustment

Exercise 58: Factor Fit Calibration

Exercise 59: Vasicek model calibration

PD IFRS 9
Module 24: Lifetime PD

PD lifetime modeling

Exogenous Maturity Vintage

Age Period Cohort

Classic Monte Carlo simulation

Quantum Monte Carlo Simulation

Quadratic acceleration over the classical Monte Carlo simulation

PD lifetime modeling

Monte Carlo Markov Chain MCMC

Quantum enhancement in MCMC

Exercise 60: Lifetime PD IFRS 9 estimation using quantum enhancements

LIFETIME PD QUANTUM
Module 25: LGD in Retail Portfolios and IRB Companies

Impact of COVID-19 on LGD

definition of default

moratoriums

Renovations and restructuring

Default Cycle

Real Default Cycles

Expected Loss and Unexpected Loss in the LGD

LGD in Default

Default Weighted Average LGD or Exposure-weighted average LGD

LGD for performing and non-performing exposures

Treatment of collaterals in the IRB

Workout Focus

Techniques to determine the discount rate

Treatment of recoveries, expenses and recovery costs

Default Cycles

recovery expenses

Downturn LGD in consumer portfolios

Downturn LGD in Mortgages

LGD in consumption

LGD in Mortgages

LGD in companies

LGD for portfolios with replacement

Module 26: Econometric and Machine Learning Models of the LGD

Advantages and disadvantages of LGD Predictive Models

Forward Looking models incorporating Macroeconomic variables

Parametric and non-parametric models and transformation regressions

Typology of LGD Multivariate Models

Linear regression and Beta transformation

Linear Regression and Logit Transformation

Linear regression and Box Cox transformation

Logistic and Linear Regression

Logistic and nonlinear regression

Censored Regression

Generalized Additive Model

Beta regression

Inflated beta regression

Support Vector Regression

Support Vector Classification

Random Forest Regression

XGBoosting Regression

neural networks

deep learning

Exercise 61: LGD econometric models

Exercise 62: Machine Learning and Deep Learning Models of LGD

Exercise 63: Comparison of the performance of the models using Calibration and precision tests.

Module 27: LGD for IFRS 9

Comparison of IRB LGD vs. IFRS 9

Impact on COVID-19

IFRS 9 requirements

Probability Weighted

Forward Looking

IRB LGD adjustments

Selection of Interest Rates

Allocation of Costs

floors

Treatment of collateral over time

Duration of COVID-19

LGD PIT modeling

Collateral Modeling

LGD IFRS 9 for portfolio companies

LGD IFRS 9 for mortgage portfolio

LGD IFRS 9 for corporate portfolios

credit cycle

Tobit Regression

IFRS 9 LGD using LASSO Regression

Machine Learning Models

Support Vector Machine

Neural Networks

Exercise 64: Estimation and adjustments for LGD IFRS 9 using Tobit regression in R

Exercise 65: LGD estimation using Bayesian Neural Networks in R

Exercise 66: LGD estimation IFRS 9 SVM

Exercise 67: LGD estimation IFRS 9 NN

LGD IFRS 9
Module 28: Advanced EAD and CCF modeling

Impact of COVID-19 on credit lines

Guidelines for estimating CCF

Guidelines for Estimating CCF Downturn

Temporal horizon

Transformations to model the CCF

Approaches to Estimating CCF

Fixed Horizon approach

Cohort Approach

Variable focus time horizon

Econometric Models

Beta regression

Inflated beta regression

Fractional Response Regression

Mixed Effect Model

Machine Learning Models

neural networks

SVM

Intensity model to measure the withdrawal of credit lines

Exercise 68: CCF OLS Regression Model in Excel

Exercise 69: CCF Logistic Regression Model in Python

Exercise 70: Neural Networks and SVM CCF in R

Exercise 71: Comparison of the performance of EAD models

Module 29: Prepayment

Prepaid and other options

IFRS 9 requirements

Probability Weighted

Forward Looking

IFRS 9 prepayment modeling

Cox regression

Logistic regression

Survival rate estimate

Joint Probability Model with PD Lifetime

Exercise 72: IFRS 9 prepayment model for mortgage in R and Excel

Module 30: EAD Lifetime for Lines of Credit

Impact of the pandemic on the use of credit lines

Lifetime measurement in credit cards

Lifetime EAD

IFRS 9 requirements

Probability Weighted

Forward Looking

Adjustments in the EAD

Interest Accrual

CCF PIT Estimate

CCF Lifetime Estimate

EAD lifetime modeling

Model of the use of credit lines with macroeconomic variables

Credit card abandonment adjustment

EAD Lifetime model for pool of lines of credit

vintage model

Chain Ladder Approach

Exercise 73: Econometric model of credit line usage in R

Exercise 74: EAD Lifetime model for individual line of credit

Exercise 75: Vintage EAD Lifetime Model for Pool of Credit Lines in R and Excel

EAD IFRS 9
STAGE DESIGN

Module 31: Preparation of Econometric Models

Review of assumptions of econometric models

Review of the coefficients and standard errors of the models

Model reliability measures

Error management

not normal

heteroscedasticity

Outliers

autocorrelation

Using Correlation to detect bivariate collinearity

Detection of multivariate collinearity in linear regression

Detection of multivariate collinearity in logistic regression

Exercise 76: Non-stationary series detection, cointegration and outliers

Exercise 77: Measurement of collinearity,

Exercise 78: Measurement of heteroskedasticity

Exercise 79: Measuring Serial Autocorrelation

STRESS TESTING
Module 32: Modernization of macroeconomic dynamics using Deep Learning

Macroeconomic models

Neoclassical growth model

Partial differential equations

DSGE Stochastic Dynamic General Equilibrium Models

Deep learning architectures

Reinforcement Learning

Advanced Scenario Analysis

Exercise 80: Bellman equation macroeconomic model using neural networks

Module 33: Forecasting PD and LGD

Trading strategies with forecasting models

Multivariate Models

VAR Autoregressive Vector Models

ARCH models

GARCH models

GARCH Models Multivariate Copulas

VEC Error Correction Vector Model

Johansen's method

Machine Learning Models

Supported Vector Machine

neural network

Multivariate Adaptive Regression Splines

Development and validation base

deep learning

Recurrent Neural Networks RNN

Elman's Neural Network

Jordan Neural Network

Basic structure of RNN

Long short term memory LSTM

temporary windows

Development and validation sample

Regression

Sequence modeling

Time series analysis with Facebook Prophet

Prediction of the spread of Covid-19

PD Machine Learning

Exercise 81: PD Forecasting with Random Forest

Exercise 82: PD Forecasting with Neural Networks

Exercise 83: PD Forecasting with Lasso Regression

Exercise 84: PD Forecasting using Recurrent Neural Networks in Python

PD Probabilistic Machine Learning

Exercise 85: PD Forecasting Gaussian Process

Exercise 86: PD Forecasting Bayesian Additive Regression Trees

Exercise 87: PD Forecasting with Bayesian Support Vector Machine

Exercise 88: PD Forecasting using Bayesian Neural Networks

Exercise 89: Forecasting PD using Pandemic variables and climate change using RNN LSTM in Python

Exercise 90: Charge-off model with VAR and VEC

Exercise 91: Charge-off model with RNN LSTM

AI STRESS TESTING
Module 34: Stress Testing PD and LGD

Temporal horizon

Multi-period approach

Data required

Impact on P&L, RWA and Capital

Machine Learning Models

Probabilistic Machine Learning Models

Macroeconomic Stress Scenarios in consumption

Expert

Statistical

regulatory

PD Stress Testing:

Credit Portfolio View

Multiyear Approach

Reverse Stress Testing

Rescaling

Cox regression

Stress Testing of the Transition Matrix

Approach Credit Portfolio View

credit cycle index

Multifactor Extension

LGD Stress Testing:

LGD Downturn: Mixed Distribution Approach

PD/LGD Multiyear Approach modeling

LGD stress test for mortgage portfolios

Stress Testing of:

Defaults

Charge Off

Net Charge Off

Roll Rates

Rating/Scoring transition matrices

Delinquency bucket transition matrices

Recovery Rate and LGD

Losses on new impaired assets

Losses on old impaired assets

Exercise 92: Stress Testing PD in Excel and SAS multifactorial model Credit Portfolio Views

Exercise 93: Stress Testing PD in SAS Multiyear Approach

Exercise 94: Stress test of PD and Autoregressive Vectors

Exercise 95: Stress Test LGD

Exercise 96: Joint Stress Test of the PD and LGD

Module 35: Stress Testing in corporate portfolios

Temporal horizon

Data required

Main Macroeconomic variables

Impact on P&L, RWA and Capital

ASRF model

Creditmetrics model

Using Transition Matrices

Use of the credit cycle index

Default forecasting

Stress Test Methodology for corporate portfolios

Impact on RWA and Capital

Exercise 97: Stress Testing PD and corporate portfolio transition matrices using transition matrix and ASRF model in SAS, R and Excel

STRESS TESTING MODELS
Module 36: ECL IFRS 9 Stress Testing

Stress testing IFRS 9 and COVID-19

Pandemic scenarios applied to the ECL calculation

Stress Testing of IFRS 9 parameters

EBA Stress Testing 2023

Treatment of the moratorium

Possible regulatory scenarios

Impact on P&L

PIT starting parameters

PIT projected parameters

Calculation of non-productive assets and impairments

Changes in the stock of provisions

Changes in the stock of provisions for exposures phase S1

Changes in the stock of provisions for exposures phase S2

Changes in the stock of provisions of exposures phase S3

Sovereign Exposure Impairment Losses

Impact on capital

Internal Stress Testing Model for ECL IFRS 9

Stage Migration

Stage Transition Matrix

Stress testing of PDs and credit migrations

Stress testing of exhibitions

Stress testing of recoveries

Exercise 98: Internal global exercise of ECL Stress Testing in R and Excel

STRESS TESTING ECL IFRS 9
Module 37: Quantum Stress Testing

Quantum economics

Classic Monte Carlo simulation

Quantum Monte Carlo

Coding Monte Carlo problem

Breadth Estimation

Acceleration applying the amplitude estimation algorithm

DGSE model using neural networks

Quantum Monte Carlo Simulation vs Normal Monte Carlo Simulation

Exercise 99: DGSE model using deep learning

Exercise 100: Quantum Monte Carlo Simulation vs. Classical Monte Carlo Simulation

QUANTUM STRESS TESTING
Module 38: Increase in Credit Risk SICR

Significant increase in credit risk (SICR)

Impact of COVID-19 on the increase in SICR risk

Recommendations Basel, EBA, ESMA, IFRS

Qualitative and quantitative criteria based on COVID-19

Increase in collective credit risk

Individual IFRS 9 credit risk increase

Phase migration matrices

Roll rate models

Markov model

Impact of COVID-19 on migrations

Estimation of PD Lifetime and PD Origination thresholds

Rating Variation

Determination of thresholds

KRIs for retail, mortgages and corporate

Increase in collective IFRS 9 credit risk

Use of discriminant test

ROC curve

false alarm rate

target hit rate

S2 size

Exercise 101: Estimation of SICR credit risk increase using ROC discriminant power test in R and Excel

INCREASED RISK SICR IFRS 9
Module 39: Models of Lifetime Expected Credit Losses

Macroeconomic scenarios impacted by COVID-19

Lifetime Loss Forecasting using macroeconomic variables

Global Exercise 102: Estimation of Lifetime Expected Loss Provisions of a consumer credit portfolio in R and Excel with VBA:

Definition of macroeconomic scenarios COVID-19

Impact of the scenarios on the estimate for COVID-19

LGD modeling using economic scenarios

CCF modeling using economic scenarios

Abandonment Modeling

Prepayment Modeling

PD PIT modeling with economic scenarios

Lifetime PD Modeling

Estimate of financial income

Cash flow modeling

Estimated survival rate

12-month ECL Expected Loss Estimate

12-month ECL estimate COVID-19 effect

ECL Lifetime Expected Loss Estimate IFRS 9 COVID-19

Stress Testing of credit risk losses

Assignment analysis of the 3 stages

Comparison of ECL Estimates

Interpretation of results in the scorecard

IFRS 9 Expected Credit Loss Validation

Module 40: ECL Validation

Initial validation

periodic validation

monitoring

Main milestones of qualitative validation

Data quality

Default Definition

Relevance of the qualification process

Override Analysis

environmental dynamics

user test

Main milestones of quantitative validation

Samples used for validation purposes

Discriminating Power

population stability

Characteristic Stability

concentration analysis

Staging analysis

Parameter Calibration

ECL backtesting

Principle 5 – Validation

IFRS 9 ECL MODEL