Optimización del Riesgo Crediticio: Predicción de Morosidad a través de Machine Learning
Cargando...
Fecha
Autores
Título de la revista
ISSN de la revista
Título del volumen
Editor
Resumen
El objetivo de este proyecto de investigación es crear modelos de predicción de clientes morosos utilizando técnicas de aprendizaje automático para que las entidades financieras tengan la capacidad de reducir el riesgo de morosidad. Los préstamos morosos plantean un grave problema para el sector financiero, ya que pueden provocar pérdidas considerables. En el caso de la empresa de estudio, se observa que solo se está recaudando aproximadamente el 62% del total esperado de las cuotas mensuales. De los 8.000 millones de pesos COP que se proyecta recaudar cada mes, únicamente se están recibiendo 5.000 millones de pesos COP. Se recopilarán y procesarán datos históricos de los clientes, tanto financieros como personales, con el fin de conocer los factores más importantes con respecto al incumplimiento. Existen varias técnicas de aprendizaje automático, como árboles de decisión y redes neuronales, se realizarán y se evaluará su eficacia en función de medidas de precisión y sensibilidad, entre otras medidas.
The objective of this research project is to create predictive models of delinquent customers using machine learning techniques so that financial institutions have the ability to reduce the risk of delinquency. Non-performing loans pose a serious problem for the financial sector, as they can cause considerable losses. In the case of the study company, it is observed that only approximately 62% of the expected total monthly installments are being collected. Of the COP 8 billion pesos projected to be collected each month, only COP 5 billion pesos are being received. Historical customer data, both financial and personal, will be collected and processed in order to learn the most important factors with respect to default. Several machine learning techniques, such as decision trees and neural networks, will be performed and their effectiveness will be evaluated based on measures of accuracy and sensitivity, among other measures.
The objective of this research project is to create predictive models of delinquent customers using machine learning techniques so that financial institutions have the ability to reduce the risk of delinquency. Non-performing loans pose a serious problem for the financial sector, as they can cause considerable losses. In the case of the study company, it is observed that only approximately 62% of the expected total monthly installments are being collected. Of the COP 8 billion pesos projected to be collected each month, only COP 5 billion pesos are being received. Historical customer data, both financial and personal, will be collected and processed in order to learn the most important factors with respect to default. Several machine learning techniques, such as decision trees and neural networks, will be performed and their effectiveness will be evaluated based on measures of accuracy and sensitivity, among other measures.