Aplicación de algoritmos genéticos en la selección de variables para la construcción de modelos de credit scoring.

Abstract

Banking entities currently manage large databases with a wide variety of variables that describe their clients, which is why these entities seek to optimize the choice of explanatory variables for the generation of credit scoring models, which is why the application of genetic algorithms in the selection of explanatory variables, they demonstrated that they generate models with better precision, separation and association measures than models with traditional techniques for selecting explanatory variables such as the methods: stepward, backward, among others, models were built with logistic regression and random forest to demonstrate this improvement in both types of models.