| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 3.25 MB | Adobe PDF |
Advisor(s)
Abstract(s)
A análise financeira é um fator preponderante na atribuição de crédito, por parte das instituições financeiras. Os modelos de classificação de crédito surgem como forma de facilitar a análise financeira, ao classificar os clientes em categorias de pagador. A aplicação destes modelos tem sido uma prática corrente em instituições financeiras, porque permitem poupar tempo e recursos. O Grupo Nors concede crédito aos clientes, como forma de fomentar a sua faturação. Assim, a área das Contas a Receber (CaR), dos Serviços Partilhados, está encarregue da análise de crédito dos clientes e tomar as decisões de conceder ou não crédito e que montante disponibilizar ao cliente. Assim, o presente trabalho tem como objetivo construir e propôr um modelo de classificação de crédito que seja útil à tomada de decisão de crédito da área das CaR e uma ferramenta que indique o montante de crédito a disponibilizar aos clientes que solicitam crédito. Desta forma, foram criados vários modelos de classificação de crédito diferentes, construídos a partir de três métodos distintos, Análise Múltipla Discriminante, Regressão Logística e Redes Neuronais Artificiais. Também, foi criado um modelo que indica o montante de crédito a conceder a cada cliente, baseado em algoritmos genéticos. Constatou-se que as Redes Neuronais Artificiais são o método de classificação com melhores resultados para o caso de classificação de crédito do Grupo Nors e que o modelo de definição do montante de crédito de clientes pode constituir uma ferramenta útil para o Grupo, uma vez que os indicadores correntemente utilizados não estão orientados especificamente para esse efeito.
Financial analysis is an important feature concerning credit assignment, for financial institutions. Credit Scoring models arise as a way to facilitate the financial analysis needed before conceding credit, as they classify clients in different categories. Financial institutions have been applying these models since they allow them to save time and resources. Nors Group concedes credit to customers in order to increase sales. So, the department “Contas a Receber” (CaR), a receiving accounts area of the Shared Services of Nors is responsible for customer’s credit analysis and for deciding whether or not to concede credit, as well as to defines the amount of credit to concede to a customer that asks for credit. Thus, this work aims to develop and propose a credit scoring model useful to CaR area and a tool to indicate the amount of credit to concede. Therefore, different credit scoring models were developed, using and comparing three distinguished methods (Multiple Discriminant Analysis, Artificial Neural Networks, Logistic Regression). It was also developed a model using genetic algorithms that indicates the amount of credit to concede to a customer applying for credit. This work shows that the Artificial Neural Networks are the classification method giving best results for the credit classification at Nors. Also the model that defines the amount of credit to concede to a customer can become a useful tool to CaR, because the usual analytical pointers are not oriented specifically to that goal.
Financial analysis is an important feature concerning credit assignment, for financial institutions. Credit Scoring models arise as a way to facilitate the financial analysis needed before conceding credit, as they classify clients in different categories. Financial institutions have been applying these models since they allow them to save time and resources. Nors Group concedes credit to customers in order to increase sales. So, the department “Contas a Receber” (CaR), a receiving accounts area of the Shared Services of Nors is responsible for customer’s credit analysis and for deciding whether or not to concede credit, as well as to defines the amount of credit to concede to a customer that asks for credit. Thus, this work aims to develop and propose a credit scoring model useful to CaR area and a tool to indicate the amount of credit to concede. Therefore, different credit scoring models were developed, using and comparing three distinguished methods (Multiple Discriminant Analysis, Artificial Neural Networks, Logistic Regression). It was also developed a model using genetic algorithms that indicates the amount of credit to concede to a customer applying for credit. This work shows that the Artificial Neural Networks are the classification method giving best results for the credit classification at Nors. Also the model that defines the amount of credit to concede to a customer can become a useful tool to CaR, because the usual analytical pointers are not oriented specifically to that goal.
Description
Keywords
Modelos de classificação de crédito Redes neuronais artificiais Regressão logística Análise discriminante Algoritmos genéticos Plafond de crédito de clientes Credit scoring models Artificial neural networks Logistic regression Discriminant analysis Genetic algorithms Customer credit plafond