Internal rating models to measure credit risks are subject to an evolutionary process. With the Model Authoring Platform, the entire life cycle of rating model implementation and maintenance is supported on one central platform:
Rating models are graphically modeled as rule trees and decision tables. There are no inherent limits on the complexity of computation and decision logic, and they can be structured and maintained by credit rating experts without any programming knowledge.
The Model Authoring Platform provides a comprehensive test environment for rating model quality assurance. The graphical Test Editor is used to define test cases with input data and expected results. Direct test execution and rule validation by ratings experts is the decisive precondition for a high-performance test-driven approach. Only error-proof and completely tested rating models are then put into functional operations.
Complete documentation of ratings models is automatically generated from the graphical model. This documentation contains full scoring and rating logic, and corresponds to the operational rating models.
Encapsulated and executable Rating Services (Java program code) are generated from the graphical rating models. Rating Services can be accessed and utilized from any third party application. If parameters are calculated solely on the basis of statistical procedures (without expert assessment), Rating Services can be performed automatically. Risk rating processes with increased process complexity (such as expert judgement based on qualitative factors or complex approval processes) are performed in the Rating Manager.
The Rating Manager supports operational execution and administration of ratings. After modeling and successful testing, rating models are deployed to the Rating Manager. From the rating models, the Rating Manager can dynamically generate the web-based user interface elements that the credit analyst needs for the operational rating process.
Validations and calibrations are an integral part of the rating model life cycle. As part of backtesting, rating models are validated and reviewed with respect to their selectivity (for instance, annually pursuant to the Basel II IRB approach).
The Model Authoring Platform takes care of model adjustments resulting from backtesting, so the model administrator can directly manage model adjustments. Based on a previously defined test portfolio, the changes are then simulated and analyzed with respect to their impact on credit portfolio.