Year of Graduation
Ярославна Де Ротшильд
Feature Selection in Credit Scorecard Modelling
Applied Mathematics and Information Science
The paper delves into the feature selection in credit scorecard modelling. A procedure for selecting the optimal set of informative features is developed. The procedure gives a solution in the area of the maximum of the AUC quality criterion. To obtain the highest-quality scoring model, we propose a procedure for optimizing parameters, grouping characteristics, and generating new features. The existing methods for assessing the quality of scoring cards are analyzed. Computational experiments on real data were carried out in the work.