Regresión logística y redes neuronales como herramientas para realizar un modelo Scoring / Logistic regression and neural networks as tools to perform a Scoring model / Regressão logística e redes neuronais como ferramentas para realizar Scoring

Resumen Introducción. El propósito de esta investigación es analizar el riesgo crediticio de una institución financiera no vigilada por la Superintendencia Financiera de Colombia en torno a un modelo scoring que permita determinar el incumplimiento de los clientes correspondiente a su cartera de consumo. Objetivo. Confrontar el poder de pronosticación de dos modelos scoring obtenidos a través de regresión logística y red neuronal. Materiales y métodos. Los modelos se desarrollan con base en una muestra de 43.086 obligaciones correspondiente a una cartera de consumo, utilizando las técnicas estadísticas de regresión logística y red neuronal. La primera está enmarcada en el grupo de los modelos lineales generalizados, los cuales utilizan una función logit y son útiles para modelar probabilidades referentes a un evento en función de otras variables, mientras que la segunda son modelos computacionales cuyo objetivo es resolver problemas utilizando relaciones ya estipuladas y para ello utiliza una muestra base del proceso que se ampara en el éxito del autoaprendizaje producto del entrenamiento. Resultados. Para ambos modelos se logra una precisión del 71% en la base de entrenamiento y del 72 % en la base de comprobación, sin embargo, a pesar de obtener resultados similares, la regresión logística arrojó la menor tasa de malos en la zona de aceptación. Conclusión. Las dos técnicas utilizadas son adecuadas para el estudio y predicción de la probabilidad de incumplimiento de un cliente correspondiente a una cartera de consumo, lo anterior, respaldado por el alto índice de eficacia predictiva en ambos modelos.

Abstract Introduction. The purpose of this research is to analyze the credit risk of a financial institution not supervised by the Financial Superintendence of Colombia around a scoring model that allows determining the default of clients corresponding to their consumer portfolio. Objective. Confront the forecasting power of two scoring models obtained through logistic regression and neural network. Materials and methods. The models are developed based on a sample of 43,086 obligations corresponding to a consumer portfolio, using the statistical techniques of logistic regression and neural network. The first is framed in the group of generalized linear models, which use a logit function and are useful for modeling probabilities related to an event based on other variables; while the second consists in computational models whose objective is to solve problems using relationships already stipulated, employing for this purpose a base sample of the process that is based on the success of the self-learning resulting from training. Results. For both models, an accuracy of 71 % in the training base and 72 % in the testing base is achieved. However, despite obtaining similar results, the logistic regression yielded the lowest rate of bad in the acceptance zone. Conclusion. The two techniques used are suitable for the study and prediction of the probability of default of a client corresponding to a consumer portfolio; the foregoing is supported by the high index of predictive effectiveness in both models.

Resumo Introdução. O propósito de esta pesquisa é analisar o risco creditício de uma instituição financeira não vigilada pela Superintendência Financeira da Colômbia em torno de um modelo scoring que permita determinar o incumprimento dos clientes correspondentes à sua carteira de consumidores. Objetivo. Confrontar o poder de previsão de dois modelos scoring obtidos através de regressão logística e rede neuronal. Materiais e métodos. Os modelos desenvolvam-se com base em uma amostra de 43.086 obrigações correspondentes à una carteira de consumidores, utilizando as técnicas estatísticas de regressão logística e rede neuronal. A primeira está enquadrada no grupo dos modelos lineais generalizados, os quais utilizam uma função logit e são úteis para modelar probabilidades referentes à um evento em função de outras variáveis, em quanto que, a segunda são modelos computacionais cujo objetivo é resolver problemas utilizando relações já estipuladas e para isso utiliza-se uma amostra base do processo que ampara-se no sucesso do autoaprendizagem produto do treinamento. Resultados. Para ambos os modelos se consegue uma precisão do 71% na base do treinamento e do 72% na base da comprovação, mas, não obstante isso, a pesar de obterem resultados semelhantes, a regressão logística mostrou a menor taxa ruim na zona de aceitação. Conclusão. As duas técnicas utilizadas são adequadas para o estudo e previsão da probabilidade do incumprimento de um cliente correspondente à una carteira de consumidores, o que precede, respaldado pelo alto índice de eficiência preditiva em ambos os modelos.

El punto de vista de los pacientes sobre la seguridad clínica de los hospitales: Validación del Cuestionario de Percepción de Seguridad / Hospital clinical safety from the patient's point of view: Validation of a safety perception questionnaire

Background: Approximately 10 percent of hospitalized patients suffer an adverse event during their hospital stay. An important proportion of subjects also feel that they have a high risk of suffering such an event during an eventual hospitalization. Aim: To determine the perception on clinical safety among patients discharged from a hospital. Material and methods: A questionnaire about hospital safety was mailed to 1300 patients discharged from a hospital. The questionnaire was analyzed using construct validity predictive validity and Cronbach Alpha for internal consistency Results: The questionnaire was answered by 384 patients, yielding a response rate of 29 percent. Of these, 77 incomplete answers were discarded. Thirty-one subjects (10 percent) reported a possible adverse event. In 19 cases (5.8 percent), it was due to medication errors and in 19 (6.1 percent), to surgical procedures. In seven cases (2.3 percent), both errors coincided (2.3 percent). According to the predictive validity of the questionnaire, if a patient reports an adverse event, the confidence in the hospital and in the professionals is reduced (p <0.001), communication with the physician is considered inappropriate (p =0.0001) and risk perception increases (p =0.003). Unsatisfied patients are those that believe that they have higher risks of suffering a medical error (p =0.005). Conclusions: Risk perception for adverse events increases after having suffered such an event. Patient satisfaction minimizes the effects of adverse events on their confidence and attitude.

[Hospital clinical safety from the patient's point of view: validation of a safety perception questionnaire]. / El punto de vista de los pacientes sobre la seguridad clínica de los hospitales: Validación del Cuestionario de Percepción de Seguridad.

BACKGROUND: Approximately 10% of hospitalized patients suffer an adverse event during their hospital stay. An important proportion of subjects also feel that they have a high risk of suffering such an event during an eventual hospitalization. AIM: To determine the perception on clinical safety among patients discharged from a hospital. MATERIAL AND METHODS: A questionnaire about hospital safety was mailed to 1300 patients discharged from a hospital. The questionnaire was analyzed using construct validity predictive validity and Cronbach Alpha for internal consistency. RESULTS: The questionnaire was answered by 384 patients, yielding a response rate of 29%. Of these, 77 incomplete answers were discarded. Thirty-one subjects (10%) reported a possible adverse event. In 19 cases (5.8%), it was due to medication errors and in 19 (6.1%), to surgical procedures. In seven cases (2.3%), both errors coincided (2.3%). According to the predictive validity of the questionnaire, if a patient reports an adverse event, the confidence in the hospital and in the professionals is reduced (p <0.001), communication with the physician is considered inappropriate (p =0.0001) and risk perception increases (p =0.003). Unsatisfied patients are those that believe that they have higher risks of suffering a medical error (p =0.005). CONCLUSIONS: Risk perception for adverse events increases after having suffered such an event. Patient satisfaction minimizes the effects of adverse events on their confidence and attitude.