Day One - Credit scoring models in the Retail segment
Model development
- Two types of models («application scoring» and «behavioural scoring»)
- Types of variables (statistical and business classification of candidate variables)
- Preparation of data (from cleaning and filtering data to the preparation, the so-called "modelling sample")
- Theoretically available and practically tested methods of making models (from decision trees, neural networks, to logistic regression and the development of "scorecards")
- Selection of variables (different measures of predictability of candidate variables)
- Statistical measures of performance (from the discriminatory power to the overall accuracy of the model)
Model calibration
- "Product-" and "client-based" models
- Determination of "cut-off score" (from the statistical background to the search for balance between protection and result)
- Combining application and behavioural components (how to get the overall credit score)
- Credit rating in Retail?
Initial model validation
- Data validation
- Quantitative validation
- Qualitative validation
Implementing the model in the loan approval process
- "Work-flow" of loan approval (key elements)
- "Credit scoring" models and loan application processing system (APS)
- Technical and user testing of the system
Follow-up model validation and model performance management
- Out-of-time validation and the champion-challenger method
- "Life cycle" model of credit quality in the Retail segment
A practical example of the described phases in SAS software
Day Two - Credit rating models in SME/Corporate segment
Model development and calibration
- Typical model components in SME/Corporate segment (financial "scorecard", "soft facts" correction, behavioural model correction)
- Types of variables (classification according to the creation method and financial classification of candidate variables)
- Preparation of data (calculation of financial indicators and initial exploratory analysis, creation of "modelling sample")
- Development of models (selection of variables - the balance between statistical measures and economic significance)
- Calibration of rating models (from financial score to final rating)
Initial model validation
- Specifics of all types of validation in SME/Corporate segment
Implementing the model in the loan approval process
- Key elements of loan approval "work-flow" – Specifics in SME/Corporate segment
- Limits of standardization of loan approval processes in the domain of legal entities
- Internal measures of credit worthiness of clients and the system for loan application by business entities
- Technical and user testing of the loan approval system
Follow-up model validation and model performance management
- Out-of-time validation and the champion-challenger method
- Guidelines for model management in SME/Corporate segment
A practical example of the described phases in SAS software