Data mining methods linked to artificial intelligence applicable to credit risk
The Financial institutions, when properly selecting their clients, reduce their credit risk, banks use different methodologies in order to classify their clients according to the default risk they present; For this we analyze a set of personal variables as well as the financial situation of the clie...
| Autores: | , , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión publicada |
| Fecha de publicación: | 2017 |
| País: | Ecuador |
| Institución: | Universidad Central del Ecuador |
| Repositorio: | Revista FIGEMPA: Investigación y Desarrollo |
| Idioma: | español |
| OAI Identifier: | oai:revistadigital.uce.edu.ec:article/61 |
| Acceso en línea: | https://revistadigital.uce.edu.ec/index.php/RevFIG/article/view/61 |
| Access Level: | acceso abierto |
| Palabra clave: | riesgo crediticio reglas de clasificación redes neuronales competitivas optimización mediante cúmulos de partículas credit risk classification rules competitive neural networks optimization by clusters of particles |
| Sumario: | The Financial institutions, when properly selecting their clients, reduce their credit risk, banks use different methodologies in order to classify their clients according to the default risk they present; For this we analyze a set of personal variables as well as the financial situation of the client that is subject to credit. The exhaustive analysis and processing of customer information takes a long time, one reason being that the data to be analyzed are not homogeneous. This paper presents an alternative method capable of constructing, from the available information, a set of classification rules with three main characteristics: adequate accuracy, low cardinality and ease of interpretation. The latter is given by the use of a reduced number of attributes in the conformation of the antecedent. This feature added to the low cardinality of the set of rules allows to distinguish very useful patterns in the understanding of the relations between the data and to make decisions. When it comes to deciding the granting of credits, it is extremely useful to have a tool of this type. The simpler the model, the smaller the number of characteristics of the subject of credit that must be analyzed so that decisions can be taken more quickly. This allows the method to be attractive to credit officers in financial institutions, since It´s possible to give a response to the applicant of the credit in less time obtaining a competitive advantage. The proposed methodology has been applied to two databases known in the literature and to two real databases of Ecuadorian financial institutions, a Savings and Credit Cooperative and a Bank that grant different types of loans and have agencies in the coast, Sierra and oriente. The results obtained have been satisfactory. Finally, the conclusions are presented and future lines of research are suggested. |
|---|