Application of Machine Learning in Financial Credit Risk Management: A systematic review

• Autor: Hermitaño Castro, Juler Anderson
• Formato: artículo
• Fecha de Publicación: 2022
• Institución: Universidad de Lima
• Repositorio: Revistas - Universidad de Lima
• Lenguaje: español
• OAI Identifier: oai:revistas.ulima.edu.pe:article/5898
• Enlace del recurso: https://revistas.ulima.edu.pe/index.php/Interfases/article/view/5898
• Nivel de acceso: acceso abierto
• Materia: machine learning; ML; management; risk; credit; algorithm; gestión; riesgo; crédito; algoritmo

Banking risk management can be divided into the following typology: credit risk, market risk, operational risk, and liquidity risk, the first being the most important type of risk for the financial sector. This article aims to show the advantages and disadvantages of implementing Machine Learning algorithms in credit risk management. A systematic literature review was carried out with the PICo search strategy, and 12 articles were selected. The results show that credit risk is the most relevant. In addition, some of the Machine Learning algorithms have already begun to be implemented; however, some have significant disadvantages, such as not being able to explain the model's operation and are considered a black box. In this sense, it discourages implementation because regulatory bodies require that a model be explainable, interpretable and transparent. Faced with this, it has been decided to make hybrid models between algorithms that are not easy to explain with traditional ones, such as logistic regression. Also, it is presented as an alternative to using methods such as SHAPley Additive exPlanations (SHAP) that help the interpretation of these models.