Examples of Overfitting Encountered When Building Private Firm Default Prediction Models
- Topics:
- Commercial Lending
- Source:
- Donald R. Frey & Co.
FREE Registration is required
Overview: The key to building default prediction models, if they are to be incorporated into credit risk management systems, is to build the most powerful model possible subject to the constraints that it is transparent, usable, and intuitive. In this process, we must constantly be on guard for whether or not we have overfit the data. This paper presents two examples of overfitting that is encountered while building private firm models. These issues, if not detected, can reduce the usability of the model and overstated the true predictive power of the model in real credit decision-making.
(Is this item miscategorized? Does it need more tags? Let us know.)
Format: PDF | Size: 246KB | Date: Apr 2005 | Pages: 14



