Publication: Diseño de una arquitectura Big Data para la predicción de crisis en el trastorno bipolar
Loading...
Official URL
Full text at PDC
Publication Date
2017
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
El trastorno bipolar conduce, muchas veces, a períodos de baja por enfermedad y la necesidad a una atención muy cercana, generando problemas económicos, laborales, sociales y familiares. La crisis, en la mayoría de los pacientes, se puede evitar mediante la predicción temprana. El proyecto “Bip4Cast” de la Clínica Nuestra Señora de la Paz (Orden Hospitalaria de San Juan de Dios) y del grupo GRASIA-GTeC de la Universidad Complutense de Madrid, propone el diseño de una arquitectura Big Data para la generación, adquisición, almacenamiento y análisis de los datos de diversas fuentes con el objetivo de predecir las crisis del trastorno bipolar. Este Trabajo de Fin de Máster es el análisis de este tipo de arquitecturas desde un punto de vista crítico. En el trabajo se han analizado y comparado distintas opciones de arquitectura del software y se ha implementado la alternativa óptima para la resolución de los problemas derivados de un entorno Big Data en el que se enmarca el proyecto. Al no existir estándares reconocidos se analizaron las propuestas más cercanas a la normalización por lo que se escogió la propuesta del Grupo de trabajo de Big Data del Instituto Nacional de Estándares y Tecnología de los Estados Unidos. Se propone un prototipo de arquitectura con la distribución Hadoop Hortonworks donde fueron implementadas aplicaciones y herramientas según las características del problema. Las pruebas realizadas en la fase de experimentación demuestran que la arquitectura cumple con el tratamiento óptimo de los datos tanto en su diversidad, seguridad, inmutabilidad y flexibilidad, otorgando al proyecto las bases necesarias para alcanzar la fase de análisis.
Bipolar disorder often leads to periods of sick leave and a close attention, generating economic, labor, social and family problems. These crises, in most patients, are avoidable across the early prediction. This work forms part of the project “Bip4cast”. It is a developed between the Clinic of Nuestra Señora de La Paz (Orden Hospitalaria de San Juan de Dios) and the G-Tec group of the Complutense University of Madrid. This work proposes the design of Big Data Architecture whose aim is to implement the generation, acquisition, storage and analysis of data from various sources to predict bipolar disorder crises. To carry out the implementation, is necessary the use of the ecosystem Hadoop, which integrates platforms that allow the storage and distributed processing of large data volumes, which makes this platform ideal for implementing architectures that can exploit the features of Big Data. There are distributions of Hadoop that implement most of the platforms that are part of the ecosystem, in addition to their own tools. In the proposed architecture, has been implemented distribution Hadoop Hortonworks that gives us various applications and tools in the different layers of a structure Big Data. This architecture has been implemented and configured according to the characteristics of the problem. The tests realized in the phase of experimentation demonstrate that the Architecture fulfills with the ideal treatment of the information so much in his diversity, safety, immutability and flexibility. These features grant to the project the bases necessary for the aim in predicting crises in bipolar disorder.
Bipolar disorder often leads to periods of sick leave and a close attention, generating economic, labor, social and family problems. These crises, in most patients, are avoidable across the early prediction. This work forms part of the project “Bip4cast”. It is a developed between the Clinic of Nuestra Señora de La Paz (Orden Hospitalaria de San Juan de Dios) and the G-Tec group of the Complutense University of Madrid. This work proposes the design of Big Data Architecture whose aim is to implement the generation, acquisition, storage and analysis of data from various sources to predict bipolar disorder crises. To carry out the implementation, is necessary the use of the ecosystem Hadoop, which integrates platforms that allow the storage and distributed processing of large data volumes, which makes this platform ideal for implementing architectures that can exploit the features of Big Data. There are distributions of Hadoop that implement most of the platforms that are part of the ecosystem, in addition to their own tools. In the proposed architecture, has been implemented distribution Hadoop Hortonworks that gives us various applications and tools in the different layers of a structure Big Data. This architecture has been implemented and configured according to the characteristics of the problem. The tests realized in the phase of experimentation demonstrate that the Architecture fulfills with the ideal treatment of the information so much in his diversity, safety, immutability and flexibility. These features grant to the project the bases necessary for the aim in predicting crises in bipolar disorder.
Description
Máster en Ingeniería Informática, Facultad de Informática, Departamento de Arquitectura de Computadores y Automática, curso 2016-2017