• 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.100.2022.tde-08032023-134543
Document
Author
Full name
Sérgio Reinaldo Marteletto
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
São Paulo, 2022
Supervisor
Committee
Lauretto, Marcelo de Souza (President)
Fernandes, Ricardo Augusto Souza
Ferreira, Fernando Fagundes
Roman, Norton Trevisan
Title in Portuguese
Técnicas de seleção de atributos através de Random Forests: um estudo de caso para detecção de tendências em séries temporais financeiras
Keywords in Portuguese
Random forests
Aprendizado de máquina
Seleção de atributos
Séries temporais
Abstract in Portuguese
Nas últimas décadas tem havido um interesse crescente em prever o comportamento futuro dos mercados financeiros. Pesquisadores investigam esse problema modelando uma representação conveniente para os dados, as chamadas séries temporais, apesar da dificuldade de estudá-las com precisão devido aos seus padrões não lineares e não estacionários. Além disso, a questão da alta dimensionalidade, presente no conjunto de dados, reduz o entendimento das relações de dependência entre as observações. O uso de novas tecnologias em finanças, como o aprendizado de máquina, busca extrair e analisar informações sobre o preço dos ativos e fluxos de negociação em um ambiente competitivo de risco-retorno. Esse trabalho propõe a análise comparativa de técnicas modernas de seleção de atributos VSURF (Variable Selection Using Random Forests) e RFE ( Recursive Feature Elimination), a fim de reduzir a dimensionalidade na base de dados. Os resultados obtidos foram consistentes e não causaram perda da capacidade preditiva do modelo.
Title in English
Feature selection techniques using Random Forests: a case study for detecting trends in financial time series
Keywords in English
Feature selection
Machine learning
Random forests
Time series
Abstract in English
In recent decades there has been a growing interest in predicting the future behavior of financial markets. Researchers investigate this problem by modeling a convenient representation of the data, the so-called time series, despite the difficulty of studying them accurately due to their non-linear and non-stationary patterns. In addition, the issue of high dimensionality present in the dataset reduces the understanding of the dependence relationships between the observations. The use of new technologies in finance, such as machine learning, seeks to extract and analyze information about asset prices and trading flows in a competitive risk-return environment. This work proposes a comparative analysis of modern feature selection techniques VSURF (Variable Selection Using Random Forests) and RFE (Recursive Feature Elimination), in order to reduce the dimensionality present in the dataset. The results obtained were consistent and did not cause predictive ability of the model.
 
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.
Publishing Date
2023-05-19
 
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.