Artificial Intelligence has quickly entered in the financial services industry covering a wide range of applications.This work studies a structured statistical approach to classify non-performing unsecured commercial exposures according to their recovery potential, based on a Machine Learning technique known as Radom Forest.The framework adopted is based on two different components: one identifying the cases that may be recovered and the jmannino.com other estimating their recovery level.In addition, the work compares the RF - introduced with a review of the underlining Decision Tree theory and its performance metrics - to the better-known Logit approach.
The framework is meant to provide an evaluation of recovery at aggregate level, for pricing and management purposes, but it is also successfully tested in comparison between two portfolios, one of which know to the analyst.Results show that the Random Forest approach is as reliable and slightly more performing than the better known Logistic approach, even with grand love red heart reposado tequila a limited set of information.