Bibliographic citations
This is an automatically generated citacion. Modify it if you see fit
Ramirez, V., (2022). Predicción del siniestro de vehículos particulares en una compañía de seguros [Universidad Nacional Agraria La Molina]. https://hdl.handle.net/20.500.12996/5576
Ramirez, V., Predicción del siniestro de vehículos particulares en una compañía de seguros []. PE: Universidad Nacional Agraria La Molina; 2022. https://hdl.handle.net/20.500.12996/5576
@misc{renati/245820,
title = "Predicción del siniestro de vehículos particulares en una compañía de seguros",
author = "Ramirez Navarro, Vanessa Judith",
publisher = "Universidad Nacional Agraria La Molina",
year = "2022"
}
Title: Predicción del siniestro de vehículos particulares en una compañía de seguros
Authors(s): Ramirez Navarro, Vanessa Judith
Advisor(s): Chue Gallardo, Jorge
Keywords: Vehículos; Siniestros; Accidentes; Seguros; Estrategias de control; Compañías de seguros; Empresas privadas; Riesgo; Modelos de simulación; Consumidores
OCDE field: https://purl.org/pe-repo/ocde/ford#4.05.00
Issue Date: 2022
Institution: Universidad Nacional Agraria La Molina
Abstract: Tras el incremento de siniestros vehiculares de los últimos años en los clientes de una empresa aseguradora peruana, se decidió desarrollar diversos modelos estadísticos que permitan identificar aquellos vehículos que generen un siniestro a futuro y cuáles son los factores relevantes asociados a la siniestralidad. Inicialmente se muestra el proceso de tarificación con el cual la empresa comenzó a ofertar el producto de seguro vehicular para luego ser reemplazado con el modelo estadístico escogido y utilizarlo como factor principal en la definición de precios de las primas. En el presente trabajo de suficiencia profesional se muestra las fases de desarrollo de los modelos Logístico y Random Forest, así como el uso de la matriz de confusión para evaluar las métricas de sensibilidad y especificidad, con la finalidad de elegir el modelo que presente la mejor predicción con respecto a la ocurrencia del siniestro. Para la implementación de los modelos se muestra la metodología Cross Industry Standard Process for Data Mining, el cual sirve para asegurar la planificación y cumplimiento de las fases establecidas en los proyectos analíticos. También se muestra los factores relevantes que se incluyeron en los modelos y presentaron una asociación a la variable respuesta, como información demográfica, financiera y de manejo del propietario; y por parte del vehículo, información de sus características como valor comercial, asientos, tipo de vehículo, colore, entre otros. Finalmente se presenta la implementación de los modelos en el negocio y el impacto positivo en los resultados de distintas frentes de la empresa aseguradora. Las herramientas utilizadas para la preparación, construcción y despliegue de los modelos fueron en la plataforma de Google Cloud Platform con el software Python.
Following the increase in vehicle claims in recent years among the clients of a Peruvian insurance company, it was decided to develop various statistical models to identify those vehicles that generate a claim in the future and which are the relevant factors associated with the claims rate. Initially, the pricing process with which the company began to offer the vehicle insurance product is shown, to be later replaced with the chosen statistical model and used as the main factor in the definition of premium prices. This work of professional sufficiency shows the development phases of the Logistic and Random Forest models, as well as the use of the confusion matrix to evaluate the metrics of sensitivity and specificity, with the purpose of choosing the model that presents the best prediction with respect to the occurrence of the claim. For the implementation of the models, the Cross Industry Standard Process for Data Mining methodology is shown, which is used to ensure planning and compliance with the phases established in the analytical projects. It also shows the relevant factors that were included in the models and presented an association to the response variable, such as demographic, financial and driving information of the owner; and on the vehicle side, information on its characteristics such as commercial value, seats, type of vehicle, color, among others. Finally, the implementation of the models in the business and the positive impact on the results of different fronts of the insurance company are presented. The tools used for the preparation, construction and deployment of the models were in the Google Cloud Platform with Python software.
Following the increase in vehicle claims in recent years among the clients of a Peruvian insurance company, it was decided to develop various statistical models to identify those vehicles that generate a claim in the future and which are the relevant factors associated with the claims rate. Initially, the pricing process with which the company began to offer the vehicle insurance product is shown, to be later replaced with the chosen statistical model and used as the main factor in the definition of premium prices. This work of professional sufficiency shows the development phases of the Logistic and Random Forest models, as well as the use of the confusion matrix to evaluate the metrics of sensitivity and specificity, with the purpose of choosing the model that presents the best prediction with respect to the occurrence of the claim. For the implementation of the models, the Cross Industry Standard Process for Data Mining methodology is shown, which is used to ensure planning and compliance with the phases established in the analytical projects. It also shows the relevant factors that were included in the models and presented an association to the response variable, such as demographic, financial and driving information of the owner; and on the vehicle side, information on its characteristics such as commercial value, seats, type of vehicle, color, among others. Finally, the implementation of the models in the business and the positive impact on the results of different fronts of the insurance company are presented. The tools used for the preparation, construction and deployment of the models were in the Google Cloud Platform with Python software.
Link to repository: https://hdl.handle.net/20.500.12996/5576
Note: Universidad Nacional Agraria La Molina. Facultad de Economía y Planificación. Departamento Académico de Estadística e Informática
Discipline: Estadística e Informática
Grade or title grantor: Universidad Nacional Agraria La Molina. Facultad de Economía y Planificación
Grade or title: Ingeniero Estadístico Informático
Juror: Valencia Chacón, Raphael Félix; Vargas Paredes, Ana Cecilia; Miranda Villagómez, Clodomiro Fernando
Register date: 30-Dec-2022
This item is licensed under a Creative Commons License