Thesis CREACIÓN DE MODELOS PREDICTIVOS PARA LA DEMANDA DE CLIENTES DE UNA EMPRESA DEL RUBRO TI
Date
2017
Authors
RODRÍGUEZ GACITÚA, IGNACIO ANDRÉS
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
El presente trabajo se centra en un proyecto de inteligencia de negocios con la finalidad de poder predecir la demanda de los distintos segmentos de clientes que posee Ki Teknology, una empresa dedicada a la creación de soluciones digitales.Para la realización de este proyecto se utilizaron tecnologías e infraestructura de big data, específicamente Apache Spark, con las cuales se crea una arquitectura capaz de procesar y administrar grandes cantidades de datos, permitiendo la creación de modelos predictivos flexibles e independientes de la cantidad de datos.Los resultados obtenidos corresponden a cinco modelos distintos, los cuales se construyeron principalmente bajo las técnicas de Gradient Boosted Tree y Decision Tree, debido a que éstos presentaron los mejores rendimientos, en base a las métricas propias de las técnicas de regresión.
This final work is a business intelligence project in order to make a prediction of the Ki Teknology client’s demamads. To do this we used the historical data of the Company related with the behavior of theirs clients.Big data tool are used to accomplish the goals of this Project, in particular Apache Spark was used as the main framework. This feature makes this work flexible and capable to process big amount of data without a problem.The results are 5 differents models. Those models were built with this two algorithms Gradient Boosted Tree and Decision Tree. Those models have a high quality predictions, that is very useful and will be very helpful for the decisionmaking
This final work is a business intelligence project in order to make a prediction of the Ki Teknology client’s demamads. To do this we used the historical data of the Company related with the behavior of theirs clients.Big data tool are used to accomplish the goals of this Project, in particular Apache Spark was used as the main framework. This feature makes this work flexible and capable to process big amount of data without a problem.The results are 5 differents models. Those models were built with this two algorithms Gradient Boosted Tree and Decision Tree. Those models have a high quality predictions, that is very useful and will be very helpful for the decisionmaking
Description
Catalogado desde la version PDF de la tesis.
Keywords
APACHE SPARK , BIG DATA , KI TEKNOLOGY , MODELOS PREDICTIVOS
