Examinando por Autor "Vega Rueda, Juan Carlos"
Mostrando 1 - 1 de 1
Resultados por página
Opciones de ordenación
- ÍtemEstimación de la probabilidad de cancelación de una póliza individual de automóviles, usando modelos de Machine Learning(Fundación Universitaria Los Libertadores. Sede Bogotá., ) Vega Rueda, Juan Carlos; Gonzalez Veloza, Jose John FredyThe purpose of this paper is to identify the characteristics of the greatest impact, which used the probability that a user will cancel an auto insurance policy. For this, it is also necessary the impact that exists due to cancellation of policies (risk of falling) and understanding the process of pricing the cost of loss of insurance. The study focuses on those individual risks, with a history of policies issued from January 2016 to December 2020. The information in the database is segmented according to A) the information of the policy: premium, sum insured for some coverages, validity periods. B) Information about the vehicle: Make, model year, type of vehicle, color, service. and finally, C) information about the insured: type of person, gender, age and levels of income or annual expenses. With the previous variables, it is intended to model the probability of cancellation using some Machine Learning models such as Random Forest and XG Boost. and the techniques of SHAP values are used to visualize and explain the impact of the most important characteristics in the models used. The results obtained show that the Random Forest model has better results, when classifying clients who make cancellations with an accuracy of 79%, and the five most important characteristics that explain the model are identified.