Developing Credit Risk Solutions using SAS

Scope

Nowadays, financial institutions face the important challenge to build effective and high-performing credit scoring systems in the context of the recently put forward Basel II requirements. It is the purpose of this 3-day course to elaborate on all steps ranging from data preprocessing to model implementation, and illustrate how they can be efficiently automated using the SAS Enterprise miner software. The course aims at providing a sound mix of both theoretical, technical insights as well as practical implementation details, illustrated by several real-life cases. It will be highly interactively organised. The course is designed for people who are involved into building scoring systems (e.g. for Basel II) and/or are responsible for monitoring their behaviour and performance.

Prerequisites

The course assumes that the participants have a basic background in statistics.

Content

Review of the Basel II Capital Accord

• The Basel II regulation
• Credit Risk versus Operational Risk versus Market Risk
• Standard approach versus IRB approach
• PD versus LGD versus EAD
• Asset Correlation
• Risk management
• Application Scoring, Behavioural Scoring, Profit Scoring, Collection Scoring, Bankruptcy Prediction
• Risk Based Pricing
• Customized scorecards versus generic scorecards
• Credit scoring, knowledge discovery in data (KDD), and data mining
• The SAS SEMMA process
• Developing scorecards

Sampling and Data Preprocessing

• Selecting the sample
• Segmentation
• Oversampling versus Undersampling
• Credit scoring characteristics
  • Application form characteristics
  • Credit bureau characteristics
• Reject inference
  • Define as bad
  • Extrapolation
  • Augmentation
• Withdrawal inference
• Determining the performance period
• Definitions of good and bad
• Direct versus indirect credit scoring
• Overriding and manual intervention
• Exploratory data analysis
• Outlier detection
• Outlier treatment
  • Truncation
• Missing values
  • Imputation procedures
  • Default values
• Nominal variables versus Ordinal variables
  • Recoding procedures
  • Classing and Binning procedures
  • Weight of evidence coding
  • Information Value
• Supervised versus Unsupervised learning

Developing PD models for Basel II

• Basic concepts of classification
• Classification techniques
  • Logistic regression (odds, p-values,...)
  • Decision trees
• Overfitting versus generalisation
• Input selection
  • Filters
  • Wrappers
  • Forward versus backward search
  • Principal component analysis
  • Multicollinearity
  • Interactions
• Setting the cut-off
  • Symmetric cut-offs, marginal good-bad rates, equal sample probabilities
• Score Points Scaling and points allocation
• Measuring scorecard performance
  • Single Sample
  • Training set/ Test set
  • Cross-validation
  • Bootstrapping
• Performance metrics
  • Classification accuracy
  • Sensitivity versus Specificity
  • Area Under the Receiver Operating Characteristic (ROC) curve
  • Cumulative Accuracy Profiles (CAP)
  • Accuracy Ratio/Gini coefficient
  • Lift charts
  • Misclassification costs
  • Gains table and trade-off charts
  • Notch difference graphs
• Defining Ratings
• Masterscale
• PD calibration
  • Point-in-Time (PIT) versus Through-the-Cycle (TTC) PD

Implementing credit scoring models

• Reporting
  • Characteristic Reports
  • Scorecard monitoring Reports
• Strategy curve
• Reweighting and realignment
• Loss Given Default (LGD) and Exposure at Default (EAD) modelling (brief)
• Profit scoring
• Tracking scorecards
• Behavioural scoring
• Survival Analysis for credit scorin