Publicação
Credit scoring as an asset for decision making in intelligent decision support systems
| Resumo: | Risk assessment is an important topic for financial institution nowadays, especially in the context of loan applications or loan requests and credit scoring. Some of these institutions have already implemented their own custom credit scoring systems to evaluate their clients’ risk supporting the loan application decision with this indicator. In fact, the information gathered by financial institutions constitutes a valuable source of data for the creation of information assets from which credit scoring mechanisms may be developed. Historically, most financial institutions support their decision mechanisms on regression algorithms, however, these algorithms are no longer considered the state of the art on decision algorithms. This fact has led to the interest on the research of new types of learning algorithms from machine learning able to deal with the credit scoring problem. The work presented in this dissertation has as an objective the evaluation of state of the art algorithms for credit decision proposing new optimization to improve their performance. In parallel, a suggestion system on credit scoring is also proposed in order to allow the perception of how algorithm produce decisions on clients’ loan applications, provide clients with a source of research on how to improve their chances of being granted with a loan and also develop client profiles that suit specific credit conditions and credit purposes. At last, all the components studied and developed are combined on a platform able to deal with the problem of credit scoring through an experts system implemented upon a multi-agent system. The use of multi-agent systems to solve complex problems in today’s world is not a new approach. Nevertheless, there has been a growing interest in using its properties in conjunction with machine learning and data mining techniques in order to build efficient systems. The work presented aims to demonstrate the viability and utility of this type of systems for the credit scoring problem. |
|---|---|
| Autores principais: | Silva, Fábio |
| Ano: | 2011 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade do Minho |
| Idioma: | inglês |
| Origem: | RepositóriUM - Universidade do Minho |
| Resumo: | Risk assessment is an important topic for financial institution nowadays, especially in the context of loan applications or loan requests and credit scoring. Some of these institutions have already implemented their own custom credit scoring systems to evaluate their clients’ risk supporting the loan application decision with this indicator. In fact, the information gathered by financial institutions constitutes a valuable source of data for the creation of information assets from which credit scoring mechanisms may be developed. Historically, most financial institutions support their decision mechanisms on regression algorithms, however, these algorithms are no longer considered the state of the art on decision algorithms. This fact has led to the interest on the research of new types of learning algorithms from machine learning able to deal with the credit scoring problem. The work presented in this dissertation has as an objective the evaluation of state of the art algorithms for credit decision proposing new optimization to improve their performance. In parallel, a suggestion system on credit scoring is also proposed in order to allow the perception of how algorithm produce decisions on clients’ loan applications, provide clients with a source of research on how to improve their chances of being granted with a loan and also develop client profiles that suit specific credit conditions and credit purposes. At last, all the components studied and developed are combined on a platform able to deal with the problem of credit scoring through an experts system implemented upon a multi-agent system. The use of multi-agent systems to solve complex problems in today’s world is not a new approach. Nevertheless, there has been a growing interest in using its properties in conjunction with machine learning and data mining techniques in order to build efficient systems. The work presented aims to demonstrate the viability and utility of this type of systems for the credit scoring problem. |
|---|