Publication: Desarrollo de un sistema de predicción de riesgo en seguros de automóviles mediante técnicas de inteligencia artificial
Loading...
Official URL
Full text at PDC
Publication Date
2019
Advisors (or tutors)
Garnica Alcázar, Óscar
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
Este trabajo de fin de grado consiste en el desarrollo de una herramienta para calcular mediante técnicas de inteligencia artificial el riesgo que tienen nuevos clientes de sufrir un accidente en sus vehículos. Principalmente, esta aplicación informática es una ayuda a las aseguradoras para intentar ajustar lo máximo posible los precios de los seguros. Las técnicas utilizadas en este proyecto han sido las redes neuronales y el Random forest. Todo el proyecto ha sido implementado en Python, tanto la parte de la interfaz grafica, con la que conseguimos que el software tenga una visión sencilla y amena, como la parte donde se realizan predicciones. El software puede ser usado por dos distintos perfiles. Un perfil técnico para personas expertas en esta área en el que pueden crear su propio modelo de predicción usando las características (Features) que crean más oportunas y un perfil funcional para usuarios que no tienen tanto conocimiento acerca de estas tecnologías. Podrán cargar un modelo ya creado y predecir en base a él. En este proyecto, se ha conseguido tener un producto con una interfaz que cumple perfectamente con las funcionalidades del software y en la parte del funcionamiento, tiene la capacidad de predecir riesgos gracias a las técnicas de inteligencia artificial mencionadas anteriormente.
This paper is aimed at the development of a tool that, through Artifcial Intelligence, calculates new clients' risk of suffering a vehicle accident. The objective of this software application is to support insurance companies in adjusting prices. For this project, neuronal networks and Random Forest techniques have been used. The whole project has been implemented in Python, the graphic interface, which allows a pleasant and simple usage of the application, and the predictor. Software is developed with two different user profiles in mind; a technical profile, for experts in this area who wish to create their own prediction model using those Features they consider the most appropriate, and a functional profile for users who don't have a great knowledge on these technologies which will allow them to upload an already existent model and obtain predictions based on it. In this project, a product with an interface that perfectly fulfils the software functionalities has been achieved. Moreover, within the operational part of the project, the software has the capacity of predicting risks through Artifcial Intelligence techniques.
This paper is aimed at the development of a tool that, through Artifcial Intelligence, calculates new clients' risk of suffering a vehicle accident. The objective of this software application is to support insurance companies in adjusting prices. For this project, neuronal networks and Random Forest techniques have been used. The whole project has been implemented in Python, the graphic interface, which allows a pleasant and simple usage of the application, and the predictor. Software is developed with two different user profiles in mind; a technical profile, for experts in this area who wish to create their own prediction model using those Features they consider the most appropriate, and a functional profile for users who don't have a great knowledge on these technologies which will allow them to upload an already existent model and obtain predictions based on it. In this project, a product with an interface that perfectly fulfils the software functionalities has been achieved. Moreover, within the operational part of the project, the software has the capacity of predicting risks through Artifcial Intelligence techniques.
Description
Trabajo de Fin de Grado en Ingeniería Informática, Universidad Complutense, Facultad de Informática, Departamento de Arquitectura de Computadores y Automática. Curso 2018/2019