Título: Arquiteturas de Redes Neurais Aplicadas a Data Mining no Mercado Financeiro Uma Aplicação para a Geração de Credit Ratings
Autores: Arraes, Daniel; Semolini, Robinson; Picinini, Ronaldo
Resumo: Credit Ratings play a central role in Credit Risk Management for Financial Institutions. Credit Risk Managers are interested in calculating Risk Spreads, in order to balance the expected losses on a portfolio of customers or credit operations, what is much easier to do using Credit Ratings. In this paper, some alternative approaches are proposed to obtain Credit Ratings, derived from Statistical Theory and Neural Networks Architectures, instead of the usual Linear Logistic Regression Model: Additive Logistic Regression, Multi Layer Perceptron Neural Networks and Bayesian Networks. A comparison of the methods is presented and some recommendations are indicated.
Código DOI: 10.21528/CBRN2001-029
Artigo em pdf: 5cbrn_029.pdf
Arquivo BibTex: 5cbrn_029.bib