Machine learning: Challenges and opportunities on credit risk

• Autor(a) principal: Costa, Patrícia Alexandra Guerreiro
• Data de Publicação: 2022
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10071/27558
• topic: Risco de crédito -- Credit risk; Data mining --; Machine learning; Linear and logistic regression; Árvore de decisão -- Decision tree; Random forest; Regressão linear e logística; Floresta aleatória

The constant challenge in anticipating the risk of default by borrowers has led financial institutions to develop techniques and models to improve their credit risk monitoring, and to predict how likely it is for certain customers to default on a loan, as well as how likely it is for others to meet their financial obligations. Thus, it is interesting to investigate how financial institutions can anticipate this occurrence using Machine Learning algorithms. This dissertation aims to demonstrate the power of Machine Learning algorithms in credit risk analysis, focusing on building the models, training them, and testing the data, and presenting the opportunities and challenges of Machine Learning that are still open to developing future studies. For this purpose, we present two Machine Learning classification algorithms: Decision Trees and Logistic Regression. In addition, numerical results obtained from various comparisons of these algorithms, which were programmed and ran in Python using the Jupyter Notebook application, are also presented. The initial sample data, consisting of 850 observations, contained credit details about borrowers in the United States of America, and is freely available data. To check the model execution and performance, between Logistic Regression and Decision Trees, we used measures such as AUC, precision and F1-score.