Análise de risco de crédito bancário com utilização da shell de sistema especialista probabilístico Spirit

• Autor(a) principal: Bueno, Tatiane de Jesus
• Data de Publicação: 2011
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Manancial - Repositório Digital da UFSM
• Texto Completo: http://repositorio.ufsm.br/handle/1/8201
• topic: Risco de crédito; Sistema especialista; Pessoas físicas; Credit risk; Expert system; Private individuals; CNPQ::ENGENHARIAS::ENGENHARIA DE PRODUCAO

The goal of this dissertation is to organize a probabilistic expert system with use of the Shell Spirit in order to evaluate the default risk of borrowers in a financial institution or minimize the risk that it represents. The study was limited to private individuals, since the variables used to analyse the risk in the concession of credit to legal entities are different, and also to the fact that this area is less explored by academics. The applied methodology was inserted in the context of a quantitative empirical research, which aimed to bring the model developed as close to reality as possible. However, in this step it was necessary to collect inside information of the institution used to study, referring to risk analysis, credit policies, profile of borrowers, and also to extract knowledge from the experts with the purpose of selecting the relevant variables for the system and to make the interaction of these when composing rules along with the definition of their weights. Consecutive to the conclusion of the model, tests occurred with some favorable situations and/or unfavorable to the granting of credit, considering that in this phase were instantiated the pertinent variables to each situation at hand. The main idea of the system is to manage and reduce the credit risk of bank institutions, for SPIRIT, an expert system is able to work with uncertainties and manipulates data, being fed with information that indicates the no default probability. The results obtained with the tests were satisfactory as they were able to identify the probability of a borrower turn out to be a no default or minimize the risk, even before the credit was released.