• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Master's Dissertation
DOI
https://doi.org/10.11606/D.55.2018.tde-06032018-110702
Document
Author
Full name
Hélio Diniz
Institute/School/College
Knowledge Area
Date of Defense
Published
São Carlos, 1999
Supervisor
Committee
Carvalho, André Carlos Ponce de Leon Ferreira de (President)
Andrade Filho, Marinho Gomes de
Thomé, Antonio Carlos Gay
Title in Portuguese
Integração de Redes Neurais Artificiais & Métodos Estocásticos para Previsão de Séries Temporais
Keywords in Portuguese
Não disponível
Abstract in Portuguese
Esta dissertação investiga a possibilidade de integração de Redes Neurais Artificiais (RNAs) e Método Estocásticos para previsão de séries temporais. O problema de previsão é geralmente abordado através de Métodos Estocásticos. Ultimamente, as RNAs têm sido muito utilizadas para a construção de previsores não lineares em diferente áreas de aplicações. Contudo, as arquiteturas da RNAs devem também ser parcimoniosas, ou seja, apenas considerar as entradas mais relevantes para realizar uma boa previsão. Assim, várias abordagens vêm sendo propostas para melhorar o projeto de arquitetura em problemas de previsão. Alguns exemplos destas abordagens são a combinação de RNAs e métodos Box 8c Jenkins, as técnicas de seleção usando métodos de poda de RNAs e modelos de RNAs com capacidade de processamento temporal. Além disso, as vantagens particulares dos previsores construídos seguindo tais abordagens podem ser combinadas através de comitês ou combinadores de previsão. Os experimentos desta dissertação foram realizados com dados sobre séries temporais de cotação de moedas e ações.
Title in English
Not available
Keywords in English
Not available
Abstract in English
This work investigates the potential integration of Artificial Neural Networks (ANNs) and Stochastic Methods for time series prediction. The prediction problem is usually solved through stochastic methods. Recently, ANNs have been used in order to create nonlinear predictors in different arcas of application. However, the ANNs architecturcs should also be parsimonious, e. g., they should just consider the most relevant inputs so as to earry out good predietions. Therefore, severa] approaches have been proposed in order to improve the architecture design in this realm. Some examples of those approaches are the combination of ANNs the Box & Jenkins method, the variable seleetion teehniques using pruning methods and ANNs dynamic models with processing temporal skills. Besides, the particular advantages of each individual predictor that are created following those approaches can bc combined through a forecasting committee. The experiments of this dissertation were carried out using real-world data sets of exchange rate and stock markets time series.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
HelioDiniz.pdf (4.99 Mbytes)
Publishing Date
2018-03-06
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.