Construcción de un score de crédito de aprobación para un sector económico de una institución financiera ecuatoriana

Abstract

In this work a methodology was developed to construct an approval credit score for consumer credits acquired in commercial firms with the financial institution who has an agreement, through generalized linear models (GLM) considering the heterogeneity of the information, which is made discretizing the explicative continuous and discrete variables through the application of the generalized additive models. These models generalize the classic models because they give a larger variety of distributions to the response variable, and no necessary the normal as in most of the classic cases; but rather is possible to try the adjust of a model with different distributions without the need of realize some modifications on the data. Besides, they also allow establishing a relationship between the dependent variable and the linear combination of predictors through an adequate function of the expected response variable. Therefore, the generalized linear model developed in this study improves the predictive ability of the good/bad client type, based on the heterogeneity of the information given by the discretization of continuous and discrete variables, which contribute to generate new credit policies, and thus, to reduce the bad portfolio.