Credit scoring statistical models
Aims
The aim of this course is to create a Scorecard for estimating probability of default. The course covers predictive modelling using the SAS/STAT software, with particular emphasis on Proc LOGISTIC. The course also deals with the Credit Scoring tool provided within the Enterprise Miner software.
Who should attend
Statistical analysts, data mining experts, business users; the topics covered will mainly make reference to the area of credit risk analysis.
Prerequisites
Basic experience in the use of the SAS language and in scoring multivariate analysis statistical techniques is required. Fundamentals in credit risk concepts should also be known.
Course outline
Brief reference to the New Basel II Accord and its impact on Credit Risk
- The three indicators
- Portfolio risk diversification and segmentation
- Expected Loss
- Probability of Default
- Default according to Basel II
Credit Scoring models for Acceptance and Trend Data
- Time references
- Model explanatory structure
Data treatment and setting up the appropriate data structure to create the Scorecard
- Sample window and performance window
- Defining sample and performance window
- Continuous and categorical windows
- Treatment of missing data
- Outlier observations treatment
Scorecard development using multivariate analysis statistical techniques
- Preliminary characteristics analysis
- Known Population scorecard
- Reject Inference
- Through the Door Population
Final Scorecard Validation
- Lift Chart
- ROC Chart
- Gini Index
Cut-off choice report
- Transforming probability of default into a score
- Score Distribution
- Cut-off Strategy Report and scenario simulation
Scorecard performance monitoring
- Distribution stability report
- Back-testing
Duration
The duration of the course is 3 days.