Thesis CREACIÓN DE UN MODELO PREDICTIVO PARA LA TASA DE ACCIDENTABILIDAD DE LOS VEHÍCULOS MÓVILES EN RUTA
Loading...
Date
2015
Authors
Journal Title
Journal ISSN
Volume Title
Program
Campus
Abstract
Este proyecto de titulación tiene como objetivo principal crear un modelo predictivo para la tasa de accidentabilidad de los vehículos móviles, que son monitoreados por una empresa outsourcing dedicada al control de flotas, llamada Wisetrack.El modelo fue desarrollado sobre un cluster de Hadoop, el cual permite el análisis de Big Data, y estructurado bajo un almacén de datos gestionado por Hive. Los datos fueron importados utilizando Sqoop, y una vez dentro de la plataforma, se aplicaron algoritmos de minería de datos implementados por la herramienta H2O.El trabajo mostró resultados favorables bajo dos tipos de algoritmos: Gradient Boosting Machine y Deep Learning. Y se hallaron algunos aspectos a mejorar, siendo el más importante la utilización de una herramienta ETL que permita realizar el proceso de transformación de forma automática
This degree project has as main object to create a predictive model for the vehicles accident rate, which are monitored by an outsourcing company dedicated to fleet control, called Wisetrack.The model was developed on a Hadoop cluster, which allows the Big Data analysis, and estructured under a data warehouse managed by Hive. The data was imported using Sqoop, and when it was inside in the plataform, it was applied data mining algorithms implemented by the tool H2O.The work showed good results under two types of algorithms: Gradient Boosting Machine and Deep Learning. And it was founding some aspects to improve, being the most important the using of a ETL tool whch could to make the transformation process in an automatic way.
This degree project has as main object to create a predictive model for the vehicles accident rate, which are monitored by an outsourcing company dedicated to fleet control, called Wisetrack.The model was developed on a Hadoop cluster, which allows the Big Data analysis, and estructured under a data warehouse managed by Hive. The data was imported using Sqoop, and when it was inside in the plataform, it was applied data mining algorithms implemented by the tool H2O.The work showed good results under two types of algorithms: Gradient Boosting Machine and Deep Learning. And it was founding some aspects to improve, being the most important the using of a ETL tool whch could to make the transformation process in an automatic way.
Description
Catalogado desde la version PDF de la tesis.
Keywords
ACCIDENTABILIDAD, MODELO PREDICTIVO, VEHICULOS MOVILES