Optimización del proceso de cobro en la gestión del riesgo de créditos masivos. Una comparación de técnicas de aprendizaje automático para predecir la probabilidad de pago en distintas fases de la morosidad.

Abstract

In the field of credit risk, scoring models based on logistic regression have been developed to optimize the assessment of default risk. However, these models require complex feature engineering and their accuracy suffers as delinquency progresses. This study proposes the use of machine learning techniques (XGBoost and Artificial Neural Networks) to generate scores in different delinquency segments (No Arrears, 1-30 Arrears Segment, 31-90 Arrears Segment, and All Segments). The Kolmogorov-Smirnov (KS) metric is used to assess the efficiency and predictive power of the models. To ensure the accuracy and reliability of the models, a five-step methodology is employed. It starts with the formulation of the problem, followed by the selection of a data sample and definition of the target variable, then a descriptive analysis of the data is performed to facilitate cleaning. Subsequently, the models are trained and tested, and finally, the results are analyzed and the models obtained are interpreted. The results show that both XGBoost and Artificial Neural Networks outperform logistic regression in most of the arrears segments. In the No Delinquency segment, XGBoost (63.36%) and ANN (61.84%) outperform LR (56.42%). In the 1-30 days segment, XGBoost (51.38%) and ANN (50.35%) also outperform LR (47.32%). In the 31-90 days segment, ANN (38.77%) outperforms LR (36.62%), but not XGBoost (34.47%). Finally, in the all-segments model, both XGBoost (74.05%) and ANN (73.59%) outperform LR (71.01%)